CHAPLAIN:  It was very interesting; I was dreaming I stood  On Jacob's ladder, waiting for the gates to open.  And the ladder was made entirely of diminished sevenths.  I was surprised, but not put out. Nothing  Is altogether what we suppose it to be. The Lady's not for Burning Christopher Fry

Current changes
Changes to PfaEdit (predecessor to FontForge)
Changes to the sfd format.

• 2-Oct-2007
  • Add Edit->Select->Autohintable
  • Typo in parsing cmap format=6 subtables could cause a crash (or eccentric behavior)
  • Can build fontforge so that it is a python extension (as well as embedding python in ff).

    $ ./configure --enable-pyextension
    

  • If a tt spline had a control point on top of an on-curve point we would lose track of it when generating the font.
  • Add support for the Uncode variation subtable of the 'cmap' encoding table (format=14).
  • add xdg-open to the list of "known" browsers.
  • Patch from Torsten Bronger. Got the args wrong in (python) font.glyphs.
  • Fabrice Popineau & Taco Hoekwater point out that in the featurefile output routines I was using a single backslash inside a printf string. The intent was to output a single backslash. But I actually need to quote the backslash with another to do that.
  • If a text string were extremely long it could overflow the 16 bits allowed for a width in a metrixedit column. Truncate to 32767 in that case. Still can't scroll over the whole thing, but editing it will pop up a better view which will allow it all to be displayed.
  • If there are no slots in the current encoding, then clicking on the encoding menu caused a crash.
  • Give the user control over the code pages/unicode ranges fields of the OS/2 table. Add python scripting commands to access them too.
  • When the user renames a glyph then things dependant on it should change too.
    So if any substitution (or pairwise positioning) refers to the old name, change it to refer to the new.
    If any contextual change refers to the old name change to the new.
    If any class of glyphs (kerning class, apple state machine, contextual by classes) refers to the old name change to the new.

    More debateable: if we change "f" to "u0066" then change "f.sc" to "u0066.sc" and "f_f_l" to "u0066_u0066_l" (and recurse on those name changes).

  • Guess at the meaning for GDEF glyph class component glyph:
    • Something not encoded
    • Something which is refered to by at least one other glyph (as a reference)
    • Something which is not generated by GSUB
  • Further improvements to the auto instructor.
  • Make the [Guess] button in FontInfo->Private active when BlueScale is selected. (and make it work too).
  • Oops. A font member in python was mis-named. Documented to be "loadState" was "loadStyle". Corrected to match the documentation.
  • Some of the 'maxp' checks in the validator were off by 1, so fontlint would report errors on the glyph with the maximum value.
  • Validator was checking non-composite glyphs against composite boundries.
  • Barry Schwartz points out that I should validate the PostScript Private dictionary (Check that entries have the right types, check that BlueValues: has an even number of elements, is in order, that pairs are separated by bluefuzz*2+1, not too big for bluescale etc.).
  • FF would crash when building accented glyphs from unicode data that included a space as the base glyph.
  • Torsten Bronger points out that "import" is a python keyword and can't (easily) be used as a member/method name. So change the glyph.import function to importOutlines.
  • Patch by Mathias Kilian OpenBSD needs an include file linux doesn't (or rather that linux imports in another include file) declaration statement mixup.
  • Patch by Michal Nowakowski to disambiguate some strings and to make sure some others get translated.
• 17-Sept-2007
  • Patch from Daniel Johnson. FF didn't compile --with-regular-link
  • And FF didn't compile if it had libpng but not libjpeg
  • Added some warning messages when doing AutoInstr so that if a font does not contain a private dictionary, or if none of the selected glyphs contain hints, then let the user know s/he isn't going to get very useful instructions.
• 15-Sept-2007
  • Added the ability to upload a font to the open font library. (You need to have created an account with them first).
  • Further improvements to the auto instructor.
  • When all glyphs are selected and the user auto-instructs, then FontForge will clear any legacy tables ('fpgm', 'prep' and 'cvt '). The old values will be of no use, and they just get in the way.
  • Add a Edit->Select->Select by Script command
    Make the Select by Name take a wildcard (so you could select "*.sc").
  • ApplySubstitutions still had problems. Caused crashes in a font with instructions.
  • If one used the metrics view to create a kerning lookup then: there would be no 'kern' feature in the feature list yet, and the display would randomly lose track of kern pairs and then create new ones. Very confusing.
  • Hmm. In some cases Merge Fonts didn't copy over lookup info.
    Add the ability to merge cross-font kerning pairs (in case the two fonts being merged have the same glyphs for those which are in both fonts).
  • Provide a mass glyph rename menu item.
  • The [Merge] button in Font Info->Lookups was disabled when two subtables in the first lookup were selected.
  • If a glyph ended in a suffix (one.superior, etc.), then changes to that glyph were not reflected in the fontview.
  • Add the ability to Populate a lookup from the selection (in the lookup sub-table dialogs)
    Do a better job of providing default values for single substitutions & positions.
  • In the Print/Display dlg, deleting something so that the next line was blank caused ff to crash.
  • The Encoding->Make From Font menu item should be disabled if the encoding is not custom.
  • Fixed a system specific crash in the BDF Info window.
  • Major rewrite of tilepath. Now has initial, medial, final & isolated tiles which it will use depending on the circumstances. No longer uses the clipboard, but you can draw the tiles in the dlg.
  • In Italic/Oblique fonts (those where the font has an italic angle specified) make the center in width command correct for the italic angle.
  • Round to Int didn't round anchor points.
  • The Control Point Info window could wander off the bottom of the screen. I was very carefully calculating its position -- and then not using the result.
  • InkScape stuffs bitmap images into svg files using the data URI
       data:<mimetype>,<data>
    (see IETF standard RFC 2397)

    So, I'll support this now as my only way of handling type3 style svg fonts with embedded bitmaps.

    The standard points out that URIs are limitted to 1024 bytes, but I don't pay any attention to that restriction, and will stuff arbetrary sized images into these things.

  • The code that set up python import/export filters did not put an end of list entry after the last one. Sometimes it would be NULL (correct) and things would work and sometimes it would be garbage and things would crash.
  • Drat. In the outline glyph view, the Import command didn't set the filter properly when you changed the import type.
  • Extend fontforge to keep track of errors encountered when loading 'sfnt' files, and then have fontlint complain about them.

    Also extend fontlint to do validity checking in glyphs against the 'maxp' table. (This won't matter when editing fonts, as ff will fixup the maxp values when it generates the font, but there might be some existing fonts with buggy 'maxp' values).

  • Oops. When changing the weight of a glyph and retaining the counter sizes, fontforge would push two entries onto the undo stack rather than one. (So the user would have to do Edit->Undo twice to get back to where s/he started from).
  • Added several new python methods to font.
    • find(contour) will search the font for all glyphs containing contour and return an interator on that set.
    • replaceAll(srch,rpl) Will search for the "srch" contour/layer in the font and replace it (each time it occurs) with the "rpl" contour/layer.
    • glyphs(type) returns an iterator over all glyphs in the font. If present type may be "GID" (default if omitted) or "encoding" and it indicates the order in which the glyphs will be returned.

    Added some new members to glyph:

    • encoding -- returns the glyph's encoding in the font
    • script -- returns the OpenType tag (a 4 letter string) of the script of this glyph.
• 31-August-2007
  • New traditional Chinese translation of the UI by Wei-Lun Chao.
  • Updated Vietnamese translation by Clytie Siddall.
  • Removed the old MetaFont command (which didn't work) and replaced it with a styles menu (incorporating the old Effects menu, and a command to change weight, to condense/extend, and to oblique. Added python scripting commands {font,glyph}.{changeWeight,condenseExtend}
  • Lots of new work on the truetype autoinstructor by Alexej and Michal.
  • Add a validation dialog which can be run just before generating a font.
  • Try a different algorithm for approximating splines by line segments (for drawing them in the outline glyph view and elsewhere). This one should show symmetry better.
  • Provide the ability to ask freetype to rasterize glyphs without hints in the Print/Display dlg (even if the glyphs have hints, rasterize them without).
  • Oh dear, rasterizing stroked fonts only worked if multilayer was defined.
  • Try to improve display of stroked fonts.
  • When stroking splines don't let miter joins grow excessively.
  • FF would remove instructions from glyphs that had at least two contours the first of which started with a control point.
  • Add a short cut for Hide Grid in the metrics view.
  • Python 2.5 initializes itself differently from Python 2.[34]. 2.[34] delay the init until an import happens, while 2.5 does it at start up. I just assumed that when I called Py_Initialize that it did so. This led to a crash on 2.[34] when I tried to use my types before the user had tried to import fontforge.
  • Further improvements to tile path.
  • When moving a control point in a ttf font where one side of the cp was a real point, I would allow an implicit point on the far side of the real point to become real.
  • When dragging truetype points around, adjacent implicit points would remain where they were (and not be implicit any longer).
  • FF would randomly crash after removing a lot of glyphs.
  • The knife tool would not cut a contour if it landed on a point.
  • Trying to add a stylename to the size pane of fontinfo generally caused a crash.
  • All blank lines in the display dlg had the same line spacing as that of the first character displayed in the dlg.
  • I don't think feature files where handling classes defined with '-' properly.
  • Read the "lib" structures from UFO/GLIF files into our python persistent data. (And write our persistent data out if it's a dict).
  • Add the ability to call hooks (python functions) when various fontforge events take place.
  • Make fontforge's basic types (Point, Contour, and Layer) be picklable so they can now be saved in an sfd file.
  • Oops. With the inclusion of the library check argument, ff would not compile if NODYNAMIC were set.
  • FontForge now stores the "userdata" python members into the sfd file as pickled objects. (FontForge's own types are not currently pickleable).
  • Add the ability to mark a glyph so that just before being saved its references will be unlinked and we will run remove overlap on it. This means the user can work with the references (and get the automatic updating they confer) and still not have a self-intersecting glyph in the output (think Aring, Ccedilla, Eogonek).
  • Create a fontlint script file.
  • When dumping both apple and opentype bitmap tables if there were a BDF table, then we'd get garbage for 'bloc'/'bhed'.
  • Add some user interface commands to python, and more importantly, add the ability to create menu items which will invoke python scripts, and the ability to add import/export conversion filters (again, python scripts).
  • Counter hints for LCG glyphs came out wrong if autohint had not been applied first.
  • Counter masks were not read out of sfd files properly.
  • Selecting a counter mask in Glyph Info caused FF to crash.
  • The Execute Script dlg would sometimes complain about invalid scrollbar size when it got closed.
  • AddAnchorPoint (scripting command) should be prepared to cast real args to integers.
  • Just as I needed a special "in use" pass of GSUB when reading from a TTC, so I also need an "in use" pass of the MATH table.
  • All this time and I've mapped "nonmarkingreturn" (GID=2) to Unicode+000C instead of Unicode+000D. I'm a twit.
• 08-August-2007
  • Support for MicroSoft's new 'MATH' opentype table. Element->Math Info.
  • Michal Nowakowski and Alexej Kryukov have improved truetype hinting.
  • kerning display is off in a pairwise lookup window if a y position adjustment were requested.
  • When reverting a compacted font ff thought the font was compacted even when it was not.
  • Make some effort to retain context in the fontview when changing encodings.
  • ff scripting command CopyReference() was broken.
  • Werner wants little popup windows in show att. Give him some for substitutions.
  • If you examined a kerning pair subtable (either with glyph info or from double clicking on the subtable in the lookup pane) and then pressed [OK], and then generated an afm (tfm, old-style kern table, etc.) you would lose all the kern pairs you examined.
  • Patch by Barry SCHWARTZ to feature file output on 64bit machines.
  • In freetype 2.3.5 the binary layout of the execution context changed.
    This means that if ff is compiled against a 2.3.4 library it can't debug truetype if linked with a 2.3.5 library (and vice versa).
    So put in checks to see that the library version matches what we expect.

    Also the contents of one of the structures has changed a bit, and a value that used to live in one field has moved to another (the ppem value moved from (exc)->size->metrics.x_ppem to (exc)->size->root.metrics.x_ppem.

    Well, David warned me that this might happen.

    Warn the user (rather than just silently failing) when we have a freetype version mismatch.

  • Patch from Barry SCHWARTZ. python fontforge.font.addPosSub was expecting its args inside a redundant tuple.
  • New Vietnamese translation from Clytie.
    Also some spelling corrections in the original messages.
  • Patch by Barry SCHWARTZ to lookup/subtable comparison.
  • If a metricsview was open, then creating a new glyph slot would potentially screw up memory and sometimes crashed.
  • Edit->Merge (and simplify) still have bugs.
  • Fix two memory leaks in the metrics view (one was huge).
  • If freetype failed to rasterize a glyph, then we would fall back on returning a bitmap image from our own rasterizer -- even if we asked freetype for a greymap. this caused problems when routines expected byte data rather than bit data.
  • If we were given a cff (otf) font, with a bluevalues array which ended in 0 (as [-20 0]) then we would omit the final zero ([-20]).
  • Outputting extension subtables appears to have been broken.
  • The code which was supposed to detect two glyphs with the same set of kerning pairs and use the same data for them didn't work (forcing us to get an extension table which also didn't work).
  • Add some more checks against non-BMP characters when drawing text.
  • For first characters in the kern pair dlg, when editing class 0, the [Edit] button was enabled but did nothing. Should work now.
  • ff could crash if it tried to create a popup image of the <New> line in the substitution pane of Glyph Info.
  • Horizontal Scrolling in the matrix edit widget sometimes forgot to move the column labels.
  • Add glyph name completion to lookup subtable dlg and glyph info panes.
  • addKerningClass didn't parse its arguments correctly.
  • Drat. When printing full page glyphs to a pdf file, the glyph was shifted up by almost the full page and we got a second (and incorrect) title.
  • MS proposes two new feature tags with it's new 'MATH' table.
• 23-July-2007
  • I have merged (and substantially rewritten) the Print and Display Dialogs. There is now only one menu item (Print) which (vaguely) the old Display dlg except that it can now be printed. The text area widget now supports OpenType features just as the metrics view does. It also supports ligature carets.

    The dialog is no longer modal, however it does not get updated with each change to the font (that would make moving a point around in the outline view far too sluggish), instead there is a [Refresh] button the user can press to force an update.

  • It occurred to me that fontforge's current mechanism for setting ligature carets requires that there be a ligature substitution is the exact number of components used to make the glyph. But in indic fonts ligatures are often made up out of other ligatures (I think) which means that there won't be enough caret positions.

    So I've added a Ligature Caret count to the Element->Glyph Info dialog to give the user control over it when necessary.

  • Add minimal support for applying apple state machine lookups in metrics view. Support is minimal because:
    • Apple seems to figure line breaks before doing substitution process but I do it afterward so I don't know where the line breaks are and I can't enter either the line start or line end state.
    • When I delete a glyph I delete it. Apple inserts a deleted glyph mark and then removes that later. State machines can respond to deleted glyphs, but I can't.
    • I don't try to figure out which feature,settings should be on by default. So the user must pick them out manually.
  • Add popup graphics to the Glyph Info and Lookup subtable dialogs to show substitutions (that is if there is an entry like 'smcp' a => a.sc then create a popup window showing the "a" glyph and the "a.sc" so the user can see what happens.
  • Another futile attempt at an embolden command. Element->Embolden
  • Add shortcuts to the anchor control dlg (Page Up/Page Down) to move to the next/previous glyph.
  • Michal Nowakowski has improved the truetype autoinstructor. He warns that it probably still has bugs (as what does not?) so I am leaving the old code available for now. He says it works best in "a clean (uninstructed) font with well defined blue zones and stems".
  • If we have a glyph with multiple encodings, and the secondary encodings occur after the primary ones, then the backmap will probably contain a secondary encoding, which means that when we go to load the font in we will probably notice the secondary encoding twice and forget the primary.
  • I used not to distinguish between ligature anchors and normal mark to base anchors. Unfortunately when I moved to lookups (from features) I had to introduce that distinction. But I didn't work through all the implications and have fixed a number of bugs related to that.
  • Barry SCHWARTZ complains that font info says "fontnames must" but that the cited adobe tech note only says "should". So change "must" to "should".
  • When processing class-based contextual lookups fontforge could not handle class 0 (the class containing "all glyphs not in another class") add code to do that.
  • If the user did not select a Gasp Version (note: active selection was needed, just seeing that it was correct and leaving it didn't work) then [OK] would leave the font with an invalid version and on some systems caused a crash.
  • View->Insert Glyph After didn't work well on a ligature glyph. It would insert the glyph after the first component of the ligature, not after the last -- which would make more sense.
  • Misnamed some private dict entries when loading from otf.
  • Make entry of ghost hints better.
  • Add the ability to determine whether a point is selected or not from python.
  • Add a mechanism so the user can ask fontforge to check for the existance of optional libraries.
  • Add range checks to some library routines which blindly referenced some BMP arrays with codepoints outside bmp. Broken by UCS2->UCS4 change.
  • Werner wants GotoChar to be able to switch sub-fonts in a cid keyed font. This may introduce bugs...
  • Revert glyph still wasn't working.
  • Someone complained that using a negative stroke took a very long time but produced correct results. Um. Ok. It's easy always to use the absolute value.
  • Oops. The mac uses UCS2 for filenames, so when I moved to UCS4 I should have changed the mac resource file interface.
  • When creating a mac resource file we only set the type/creator fields and failed to initialize the finderFlags.
  • Fixed crash bug in generating a cursive connection anchor sub-table.
  • In TrueType composite glyphs with the USE_MY_METRICS bit set the lock icon wasn't scrolled properly.
  • The change from UCS2 to UCS4 broke text copy/paste.
    We failed to add a terminating NUL of the right size in all cases
    We continued to use charset=UCS-2 when it should have been UCS-4
  • FF crashed when trying to View->Show ATT on a font that appears to me to contain an invalid 'kern' sub-table. I have removed the proximate cause of the crash. I have provided a warning that the kern table appears invalid. And I have cleaned up my internals after detecting the bad sub-table (I had a lookup with a feature but no script and this caused problems).
  • Multiple substitutions in the metrics view did not properly update the count of glyphs to be displayed.
  • If nothing changed in the metrics input field and the output contained a ligature (or a mult subs I suspect). FF would complain.
  • in python, font.generate() didn't work. The PyArg_ParseTupleAndKeywords behaves in a way I did not expect.
  • [Bottom] and [Down] still didn't work for lookup subtables.
  • Try to force the text field in the metricsview to a fixed size. In some fonts it seems to be initialized to a huge value.
  • Add ability to display italic side bearings in the char view.
  • The lines drawn for italic fonts to show the italic origin and width were at slightly the wrong angle (I used a sine when I should have used a tangent).
  • If a font did not have any horizontal metrics then ff would not set the em-size.
  • Show Att trampled on memory when displaying apple contextual substitution state machines.
  • Wasn't parsing apple's 'lcar' table properly.
  • if a font contained a 'post' table but didn't name all names (or something like that), then the attempt to name the glyph based on the encoding was broken after the encoding change.
  • FF did not recognize that a bdf file was greymaped. Broken by the bdf properties work a year ago or so.
  • The import lookups button in fontinfo forgot about the subtables (sort of).
  • The metricsview used the wrong count field to determine whether things changed. It used the glyph count, not the char count (which meant that when we had a ligature and the number of chars was greater than the number of glyphs, things got confused.)
  • In the metrics view, anchored attachments only worked if the base glyph were itself unmoved (that is the mark was placed relative to the unmoved location, not the actual location).
  • The search dialog should provide user with control over the error bound.
    The rotate checkbox didn't work if the flip checkbox wasn't checked.
  • If a replace contour added a control point to a point that did not have one (went from a line to a curve) then that control point would get lost.
    If a search matched across the start point of a contour then search/replace could go into an infinite loop if the search and replace paths were the same.
  • Find/Replace (replace) didn't work on quadratic splines.
  • Add two python methods:
    • Layer.interpolateNewLayer(other-layer,amount)
    • Font.createInterpolatedGlyph(glyph1,glyph2,amount)

    The first creates a new layer by interpolating between the current layer and the layer in the first argument.
    The second creates a new glyph in the font by interpolating between the first two arguments. The glyph's unicodecode point and name will be copied from the first argument (the font must not already contain this glyph).
    If amount is 0 the result will look like the first glyph, if 1 then like the second.

  • When recovering from a crash, FF would sometimes complain about a mismatched version number. Don't.
  • Problems parsing 'mort' tables could cause a crash.
  • When building a contextual lookup, don't list that lookup as something that it could invoke (ie. list all lookups in this table (GPOS/GSUB) except for ourselves). Don't want to encourage users to create infinite lookup loops.
  • Point matching didn't work when there were references to references and multiple references within a glyph.
  • FF was having problems with extension lookups with multiple sub tables.
  • We were trying to print a trailing NUL in some strings from the fontview.
  • Kerning by classes got broken in metricsview by the addition of support for device tables.
  • A GPOS contextual lookup only listed GSUB lookups in the lookup/sequence dlg
    Pressing [OK] in the lookup/sequence dlg caused a crash if no lookup selected.
  • Openfontdlg was looking at the filter listbutton rather than the rename namelist listbutton.
  • mf2pt1 now uses "glyph_dimensions" rather than "bbox"
  • The metrics view should now handle device tables.
  • Goto could crash when used on small encodings.
  • -lang wasn't permitted before -c.
  • Use numeric text fields for anchor positioning.
  • Graham Asher points out that the meanings of underline position in the 'post' table and the FontInfo dictionary are different. One refers to the top of the underline rectangle and one to the center of it.
  • Align point would crash if the selected point were the end point of a contour (or if the two points around it were in the same place).
  • The baseline was not properly located when displaying it in the fontview.
  • The scripting command BitmapsAvail would generally cause FF to crash if done when there was a UI..
  • We seem to be misimplementing my obsolete (sfd file) convention for having duplicate encodings point to the same glyph. Result was that occasionally a glyph would be removed and a pointer to something it refered to would be put in its place.
  • Change the name of activeFontInUI to activeFont
    Add an activeGlyph method.
    Add the ability to call a python script from a outline view.
  • Hmmm. If a textfield is shifted right, and then resized so there's now room for all the text, the unshift it.
  • Werner suggests that it would be useful to be able to specify wildcards in the goto dlg.
  • Michael Zedler tells me that glyphs output by mf2pt1 contain a line:

     % MF2PT1: bbox 0 90 834 422
    

    where the third (so called) bounding box entry is actually the glyph's advance width. I was reluctant to use this at first, because that clearly isn't something that belongs in a bounding box...

  • Werner tells me that lilypond uses a slightly different syntax for the MF2PT1 bbox comment, so make our parsing slightly more generous.
  • When creating a new lookup subtable for an anchored lookup, it did not get marked as having anchor classes and feature file output failed because of that.
  • When outputing single lookups, the feature, script and language tags all had ^A where they should have had the second letter of the tag.
  • Remove the code to produce the old, broken, 'size' feature.
• 07-June-2007
  • Add support for Adobe's feature file syntax.
    Use Merge Kern (now Merge Feature) to read one
    Use Element->Font Info->Lookups->popup menu->Save to generate one

    Sigh. Adobe documents a feature file syntax they don't fully support, and even their documented syntax is inadequate to represent all the complexities of OpenType.

    FontForge supports anchors and device tables and will parse lookups which contain these. This is a syntax which Adobe documents but does not implement. Adobe also warns that this syntax may change when they get around to implementing it.

    FontForge makes one extension to the syntax which allows me to specify complex contextual features.

    Neither I nor Adobe support reverse contextual substitutions, nor GDEF mark classes.

  • Add a popup menu to the lookups pane of font info.
  • Change the name of File->Merge Kern Info to File->Merge Feature Info
  • Remove the code to produce the old, broken, 'size' feature.
  • An earlier patch, designed to make the tabset work in the charview even if the user changed the name, didn't work.
  • Bug in the selection iterator.
    Also Max Rabkin points out the docs say the iterator returns selected glyphs rather than encoding slots. Fix the docs, then add a member (.byGlyphs) which will make the iterator return glyphs.
  • Accept contour += (1,2) rather than forcing people to say contour += fontforge.point(1,2)
  • Was using an error return from two setter functions (contour.set_closed, set_is_quadratic) when I should have returned success.
  • Crash when mistyping a glyph name in Glyph Info possub panes. Probably in the equivalent lookup dlgs too.
  • If a glyph contained two copies of another glyph, and one of those copies was off by enough that ff didn't recognize it, the depending on how the contours were ordered, ff might fail to recognize the first glyph too, even though it did match.
  • Add iterators to contours and layers. Python users seem to expect them even when they aren't documented to exist.
  • layer += contour didn't work in python. I was examining a bit through the wrong data type.
  • Provide a mechanism for copying lookup information from one glyph to another. (Edit->Copy Lookup Data)
  • Loading kerning info from a mac 'FOND' resource was broken by the lookup change.
  • Python scripting command font.createChar didn't. Partial patch from Max Rabkin.
  • Adding a kerning pair from the metrics view would often add two or three pairs with the same character combination and different values.
  • Another remove overlap bug. Introduced recently.
  • Reinstate Apply Substitutions scripting command.
  • Add some dutch styles.
  • Make yet another attempt to correct for rounding errors when loading a postscript font.
  • Add a way to get a font's filename to python.
  • Werner points out that in scripting I use "break" in an example. It isn't documented to work and I don't implement it but I do show it. Oops. Ok, make it work.
  • Display substitutions was broken.
  • When generating default names for subtables, it is confusing to name them the same as the lookup.
  • Change to using USC4 from UCS2 internally.
  • Provide default names for lookups once the user sets the feature.
  • Disambiguate some strings for the polish translation
  • Revert glyph should not work for old sfd files.
• 11-May-2007
  • Loading some afm files would go into an infinite loop. (Introduced 9-May)
  • [Import] button on Font Info->Lookups crashed.
  • Fix a number of problems related to anchor classes.
  • After the user does an [Add Lookup] we should select the new lookup, to make it easier for him/her to do an [Add Subtable].
• 9-May-2007
  • When ff parsed an afm file it only read the first ligature on a line.
  • The [Call...] button in the scripting dialog is meaningless for python scripts (or I don't know how to give it meaning) so disable it.
  • Vertical scrolling in the metrics view tried to scroll the list of features too.
  • Tab, and arrow keys didn't work in the matrixedit widget after I added hidden columns. Make Return do the expected thing (move to start of next line and create it if possible).
  • Patch by Pierre to fix uninitialized value when compacting encodings.
  • Typo in realloc of lookups meant that nothing happened (when it should have increased array size by 1).
  • Add the OFL icon to the fontinfo dlg.
  • Plugins were broken with regular link configure option.
  • Inkscape can start X on the mac before it tries to connect. Hmm. I can do that too. Inkscape manages to start X without an xterm. I don't know how to do that.
  • When reading PS and CharStrings, the closepath operator didn't allow for rounding errors.
  • Metrics View would crash if more than 20 glyphs entered.
  • Alexej Kryukov points out that it would be nice if moving an anchor point did not set "changed since last hinted.
  • Alexej Kryukov points out that the "changed since last hinted" flag gets set inappropriately.
  • Alexej Kryukov points out that the Anchor control menu item looked as though it had changed, but actually had not.
• 1-May-2007
  • I've redone the way FontForge handles lookups so that it is more in keeping with OpenType.

    The major change is that fontforge now presents lookups to the user rather than features. I think this makes simple things more difficult (which is why I avoided this when I started), but it makes complex things possible.

    Sadly the world is not simple.

    So when a piece of typographic information is created (a ligature, a kern pair, a glyph substitution, etc.) it must be tagged with a lookup (actually a lookup sub-table) rather than a feature tag.

    The lookup itself will be tagged with a feature tag (possibly several tags) and with scripts and languages in which that lookup should be active.
    NOTE: This reverses the way GPOS/GSUB think about things, but it contains the same information.

    The Font Info dialog now contains a Lookup pane which allows you to create and edit lookups and their subtables. You can also reorder them. The order shown in the dialog is the order in which they will be applied.

    A mac feature/setting subtable also gets converted into this format.

    The Font Info dialog no longer has Anchor Classes, Contextual, or State Machine panes.

    Instead you can edit a lookup subtable's data (for any subtable type). There are new dialogs which list all the information for each lookup type (ie. a dialog which lists all kern pairs in a subtable), and these provide access to the old anchor class, contextual or state machine dialogs.

    The Glyph Info command has also changed. It looks simpler and more comprehensible (I think), but the act of creating a new substitution has become more complex because (potentially) one must create a new lookup and lookup subtable before doing the simple task of adding a new replacement glyph.

    The kerning class, contextual and state machine dialogs have all changed in that they no longer request a feature tag, they now need a lookup subtable.

    The metrics view also needs a subtable.
    The metrics view has changed considerably.
    It displays all the features in the font, and allows you to select which ones you want active in the view. It lets you set the script & language. It will apply lookups that it couldn't handle previously like ligatures and contextuals.

    It does not do Indic glyph reordering. I'm not sure how to and last I checked MS had not updated their docs to reflect their new procedures.

    Many other dialogs now take a lookup subtable where appropriate.

    Show ATT has changed, but it is still not editable. I hope that the Lookups pane will do that instead.

    There used to be a scripting command which indicated what ligature features got stored in afm files. Now each ligature lookup has a flag set on it which conveys this information.

    The Element->Typographic Features menu has been removed. It's functionality has moved into Font Info->Lookups (I hope I've got everything).

    Some scripting commands have been removed, others have been changed and others have been added. I apologize for this, as it will break existing scripts, but some basic concepts no longer exist and others, very different, have replaced them. (and anyway people may switch to python scripting now).

    Removed:

             DefaultATT
             ControlAfmLigatureOutput
             ApplySubstitutions
             CopyGlyphFeatures
             AddATT         Replaced with AddPosSub
             RemoveATT
             GlyphInfo(Position/Pair/Substitution/AltSubs/MultSubs/Ligature)
             SelectByATT    Replaced with SelectByPosSub
    	 
    

    Changed

             Set(V)Kern takes an optional third argument, a lookup subtable
                  name (if not specified it choses one)
             AddAnchorClass(name,type,lookup-subtable-name)
             GetPosSub(subtable-name)
             AutoKern(spacing,threshold,subtable-name[,kernfile])
    	 
    

    Added

            AddLookup(name,type,flags,feature-script-lang-array[,after-lookup-name])
                 n*[feature-tag,script-lang-array]
                 n*[script-tag,lang-array]
                 n*[lang]
             GetLookupInfo(lookup-name) =>
            [type,flags,feature-script-lang-array]
             AddLookupSubtable(lookup-name,subtable-name[,after-subtable-name])
             GetLookupOfSubtable(subtable-name)
             GetSubtableOfAnchorClass(anchor-class-name)
             AddPosSub(subtable-name,variant(s))
                      (subtable-name,dx,dy,dadv_x,dadv_y)
                      (subtable-name,other-glyph-name,dx,dy...)
             RemoveLookupSubtable(subtable-name)
             RemoveLookup(lookup-name)
             MergeLookupSubtables(subtable-name1,subtable-name2)
             MergeLookups(lookup-name1,lookup-name2)
             SelectByPosSub(subtable-name,search_type)
             GetLookups("GPOS"/"GSUB")
             GetLookupSubtables(lookup-name)
             LookupStoreLigatureInAfm(lookup-name,store-it)
             LookupSetFeatureList(lookup-name,feature-script-lang-array)
    

    The sfd format has changed. New files are tagged as version 2. Old files will still work, but ff will no longer produce file in the old format.

  • New translations to match the new UI: Polish, Greek, German, French, simplified Chinese.
  • FontForge now supports python scripting if it has been configured to do so.
  • Tell users they should add their email address to the License after pressing the [Add SIL Open Font License] button.
  • The metrics view had numerous problems. Most significant if a glyph changed elsewhere the metrics view would not be updated.
  • If the width of a glyph changed in the metrics view then that was not synchronized to dependent glyphs.
  • Enforce the "Use My Metrics" bit of a reference.
    In TrueType, if this bit is set on a reference then then width of the containing glyph will be the same as that of the reference.
    So don't let the user change the width of such glyphs. (and add a lock icon to the width line in the character view to show that it is fixed).
  • Fix a number of Remove Overlap bugs.
  • FF would destroy memory in the metrics view if the user attempted to display a character which was not in the font.
  • Was setting the last char index (of OS/2) incorrectly if the font contained non-BMP characters. Might also have set the first char index wrong two.
  • Bugs in 'PfEd' table, glyph comment subtable.
  • Couple of places where I actually say "FontForge 1.0". Let's bump that to 2.0 now that I've done that for the sfd format.
  • Overlap problems with incorrect path orientations.
  • Arrow keys didn't work in the charview if the tabset were active.
  • I used to use "><" on the font's window title to represent a compacted font. No one understood it. Use "Compacted" instead.
  • Philipp Poll wants to see the side bearings (I think) in the charview.
  • Nicolas Spalinger of SIL suggests a help button next to [Add OFL] to open a brow ser on the official OFL site.
  • Patch from Nicolas Spalinger of SIL regarding the verbiage surrounding the OFL i n the fontinfo. Also his suggestion to make the tooltip live longer.
  • Replace with Reference didn't check reference offsets correctly.
  • Find problems didn't find references which intersected.
  • Add some standard classes to the context chaining dlg.
  • Make sure we only generate a postscript name in the name table once.
  • The change length button of the cvt dlg didn't work.
  • Input/Output of Apple's 'lcar' table was inconsistent. I don't know if it is "right" now, but at least ff can read back what it wrote.
  • Werner points out that the debug windows show up at odd places if the window gets resized.
• 12-Mar-2007
  • Add a button to insert the SIL Open Font License into the 'name' table.
  • FF failed to store TeX heights and depths properly in a tfm file. FF failed to read or write TeX extension lists properly.
  • Patch from Reuben Thomas to turn off the splash screen from the preference dlg.
  • I had the dotlessj code point wrong. Again.
  • FF did not center the charview window well on an empty glyph.
  • The scripting command to add instructions to a glyph could immediately lose them if the glyph's points weren't numbered properly
  • We used to allow the syntax "<number><numbers>* command" in ttf instructions (which automagically inserted an implied push). Someone noticed that

    PUSHB_1
    1
    2
    

    would lose the second number and not complain. In fixing that, I broke the first. Now both should work (I hope).

  • Code which I hope will handle German keyboards and make them do what the menu says they (ignoring the shift key) should for certain hot keys (brackets, curly brackets, brokets all get different modifier sequences on a german keyboard than an english).
  • If a glyph had no hints then ff could add a hintmask operator (which didn't try to control any hints). Now I can't find anything in the docs which SAYS it's illegal to have a hintmask in a glyph with no fonts, it does seem bad practice, and a program called tx.exe complained about it.
  • Patch by Michal Nowakowski to allow a nicer notation for delta instructions in the instruction editing window: Example:
    PUSHB_3 6@12 8 1 DELTAP2
    This moves point 8 by six delta-steps (that's clear) at 37 ppem (that's not clear at first glance - assuming delta_base is 9, DELTAP2 works from 25 ppem to 25+15 ppem, in this example it's 25+12 ppem). So it's just a way of constructing an argument byte for DELTAP/C.
  • Allow FontForge's menu shortcuts (hot keys, whatever) to be user definable. We use a gettext based mechanism, type $ make FontForge-MenuShortCuts.pot to get a pot file for the process. It needs to be converted to an mo file and installed in $(prefix)/share/locale/$(LANG)/LC_MESSAGES, just like any other getext file.
  • The Export scripting command did not work as documented. It is documented to export all selected glyphs. Instead it exported all glyphs.
  • Redo the anchor control dlg.
  • Be a little more lenient about bitmap bounding boxes for empty glyphs when loading bdf fonts
  • Make the ClearHints scripting command take arguments: "Horizontal" or "Vertical" to allow users to choose to clear either horizontal or vertical hints (the old form with no args still works and clears both, of course).
  • Typo gives us an unititialized variable when saving ufo.
  • Various problems to various anchor dialogs.
  • If an anchor point contained an x devtab but not a y devtab then it produced a pointer to garbage for the y devtab instead of NUL.

    Pairwise positioning lookups could also get their device table offsets screwed up.

  • Werner points out that if you have a series of the same glyph in a row in the encoding (generally .notdef) and you are in a char view, then pressing Next Glyph does nothing. You are stuck at the same code point. (same thing happens for Prev Glyph).
  • a chinese input method steals keystrokes I use for hot keys (I think). So if users don't want glyph navigation by typing characters in the charview (controled by a preference item) then don't turn on the input method when the window is created.
  • Add hot key for Select->Points on selected contours.
  • Assign "dotlessj" to map to 0x237, the unicode code point, rather than to 0xf6be, the old Adobe private use area. !!!! THIS DOES NOT CONFORM TO THE AGL !!!!!
  • Oops. The gloss for the shift point/contour/zone was slightly off and gave the wrong information.
  • Add vietnamese font styles (bold, italic, regular, etc.)
    From Clytie Siddall.
  • Fixes for rounding errors in finding spline intersections.
  • Werner points out that removing a bitmap glyph from the bitmapview does not force a redisplay in the fontview.
  • Gleep. Werner points out that when generating bitmap strikes, ff would always generate a temporary postscript font to rasterize, so any truetype instructions would be ignored. And the generated strikes would be ugly.
  • Add a command to show anchor point attachments within the outline character view
    (View->Combinations->Anchor Glyph at Point)
  • Force point numbers to be correct when changing glyphs in the charview.
  • Add support for Apple/Adobe's obsolete 'typ1' sfnt. A way of stuffing a type1 font into an sfnt wrapper.
  • Even though the viewbox/output width hieght ratios are different, people want the image scaled the same in svg input.
  • People would like to have only one charview and double clicking in the fontview just changes the contents of an existing window.
    Preference Item: OpenCharsInNewWindow
  • Add a tabs line to the outline char window. If users are going to have only one charview open then they probably want some history too.
  • Werner points out that what I call "Type42 CID" should actually be called either "Type11" or "CID Type 2".
• 20-Dec-2006
  • Patch by Johan Winge fixing a bug in dragging out a magnification rectangle
  • Patch from SUSE fixing a variable which can be uninitialized.
• 19-Dec-2006
  • I have received many complaints about the grey background, so I have tried various other backgrounds -- and gotten complaints on all of them. Currently try using a white background by default and add a discussion in the FAQ about changing it.
  • Add support for Unified Font Objects (a directory holding many Glyph Interchange Format files). People at the Open Font Library seem to think this is useful. Also support for importing/exporting glyphs in glif.
  • People asked me to implement the UFO/GLIF format in part because each glyph is stored in a separate file giving finer granularity to version control. I worry about using UFO/GLIF because it is not documented to store all the information needed to regenerate a font (being mainly concerned with glyph outlines). So I've implemented a directory based version of the sfd format which I hope will provide the one glyph per file format people say they want.
  • MS has extended the 'gasp' table to give users control over clear type. Version number has changed too. FF now supports the new bits.
  • MS has also added a cleartype bit to the head table.
  • Hmm. I wasn't setting the apple specific bits of the head table properly.
  • Allow the user to create an empty bitmap strike.
  • Add the ability to remove bitmap glyphs from a strike.
  • Fix a couple of memory leaks.
  • The docs don't mention this, but... mangal.ttf has a NULL offset for the lookahead class definition, so if that happens -- no classes.
  • Werner doesn't like that I always add dummy bitmaps for .notdef, .null & nonmarkingreturn. It seemed logical to me that if I were prepared to add those glyphs to the outline font I should be prepared to add them to the bitmap strike. But Werner (I think) sees bitmap strikes as containing exceptional glyphs which need special handling, whereas I just saw them as complete fonts. So in this context I guess there is no need to add these glyphs. As a compromise I have decided that I will add the glyphs to the bitmap strike if I added the glyphs to the outline font.
  • Add scripting functions:
    • SelectMoreIf
    • SelectMoreSingletonsIf
    • SelectSingletonsIf
  • The changed flag wasn't being set in Remove ATT.
  • GetPosSub scripting command trashed memory.
  • Make default char of OS/2 be space for otf and .notdef for ttf. This is a guess at what should be done.
  • FF was still setting firstcharindex in OS/2 incorrectly. The problem is that glyph 1 (or 2?) has encoding 0 in the mac cmap table while it is not referenced at all in the windows cmap.
  • Wasn't setting os2_typelinegap field when reading an sfnt.
  • In a font without any LCG characters, fontforge set the OS/2 capheight and xheight fields to -1e10. It should have been 0. Patch by Roozbeh Pournader.
  • Patch by Roozbeh Pournader. Regular bit in fsSel in OS/2 was being set when it should not have been.
  • Werner suggests that when looking at a bitmap strike, nextdef and prevdef look at the next glyph defined in the strike rather than the font as a whole. Useful for sparse bitmaps
  • Werner wants to be able to set bitmap magnification in the font view.
  • The "Recalculate Bitmaps" button in the bitmap view should adjust its positioning as the window is resized (and be sensitive to other languages which might have different button widths).
  • there was an uninitialized variable used in parsing contextual by coverage table.
  • Werner points out that when double clicking on a bitmap glyph in the font view, the glyph is created / rasterized without using freetype.
  • Sigh. Ok. make the size of the label font in the fontview configurable.

    fontforge.FontView.FontSize: 24
    

  • Werner finds "Original" confusing for an encoding and suggests "Glyph Order" instead.
  • MS has revealed the meaning of a few new GETINFO bits. Put them in the Gloss window.
  • Set/Select aren't very informative in class definition windows, change to "Set from font"/"Select in Font"
  • Improve the way glyphs are centered in the outline char view. At least I hope it's an improvement.
  • Oh. It is perfectly legal to have a context chain rule with no substitutions. This can be used to avoid matches in complex cases so:

    [a-z] a b => does nothing
    a b [a-z] => does nothing
    a b => substitutes
    

    will only substitute of "ab" is a word all to itself.

  • If the metrics view were displaying text containing non UNICODE characters then some edits in the text field which displays glyphs would leave us with the replacement character showing instead.
  • Werner wants to know whether storage locations are uninitialized or not, and to get warnings when reading an uninit value.
  • Werner wants to see IF statements indented in truetype instruction display.
  • FF wasn't telling freetype to set the mono bit when loading a glyph, which meant that GETINFO would return the wrong value for mono rasterization.
  • If there are too many numbers after an NPUSH[BW] instruction, FF silently ignores them.
  • Anti-Aliased rasters still weren't aligned properly with the outlines in the debug window.
  • When debugging I provided popup information for anti-aliased rasters. But I didn't for mono rasters.
  • ff would crash when asked to debug fpgm/prep (because my rasterization routine didn't work if there were no contours).
  • Werner wants to turn off auto goto in the charview when debugging is going on.
  • The gloss information for delta point (and cvt) was wrong if the magnitude selector was 7. I claimed it moved (changed) by 0, when it should have been by -1.
  • Werner wants to be able to set memory watch points on storage and cvt locations. Set these through the cvt/storage windows by clicking on the appropriate locations. Modify the points window so the user can set watch points by clicking on it too.
  • The Points window would sometimes scroll in unexpected ways.
  • Werner wants to see the raw value of a "WS" instruction in the gloss.
  • merge fonts would get the script lang index of merged kern pairs wrong.
  • Correct some of icc's warnings
  • Apostolos points out that NAME_MAX isn't defined on solaris, but _POSIX_NAME_MAX is. But _POSIX_NAME_MAX doesn't seem to be defined on linux.
  • Display fonts would crash if you tried to change the font. But introduced in last release.
  • Patch from Mike FABIAN. Rasterizing by freetype only initialized a variable if CONFIG_TYPE3 (multilayer) was set.
  • StrSplit & StrJoin scripting functions.
  • Patch by Rüdiger Oertel. Strncat could overflow a buffer.
  • Make simplify do point catagorization, and not just of points changed.
  • Curved stems are also limited by geometry. Can't just assume control points work.
  • Why are pdf files listed as a source of bitmap fonts?
  • When examining a bitmap font, metrics view would crash when asked to look at a non-existant glyph.
  • The list of verbs for indic glyph rearrangement didn't display.
  • Converting from the mac korean encoding didn't work.
  • Images imported into a bdf character did not set the character used bit.
  • Add a CopyAnchors command.
  • Patch from Raph Levien. FF mishandled seac if the accent were not translated.
• 26-Oct-2006
  • If a large group of glyphs were autohinted at once, and one of those glyphs contained a reference to a glyph that appeared after it in the font, then that glyph would still end up with the dirty hint mark set.
  • It was impossible to set preferences on the Accents pane. They would either always be 0 (or false), or make spurious complaints.
  • Add a Ucs4 scripting command (reverses Utf8).
  • when creating an array with [...] notation (in scripting) I forgot to dereference lvalues.
  • Add scripting commands
    • Revert
    • RevertToBackup

    Make the "+" operator concatenate arrays

      [1,2] + [3,4] == [1,2,3,4]
      [1,2] + 3 == [1,2,3]
    

  • Add a scripting command, FontImage, which generates a font thumbnail.
  • Make it possible to save the Font Display dlg as an image.
  • Clean up the Display dialog a bit. The concept of "unhinted" truetype is obsolete. Remove it.
  • At Andrey V. Panov's request I have added a scripting command "CopyUnlinked" which behaves like Copy except that any references will be unlinked (turned into splines) as they are placed in the clipboard.
  • I believe OpenType would prefer if formed arabic glyphs didn't use their unicode encodings, but used simple substitutions instead. Accept both for svg output (previously we only accepted unicode encodings).
  • Patch from debian (bug392688) from Wojciech Mu*a and K*stutis Bili*nas
• 19-Oct-2006
  • Patch from Mike FABIAN, array overflow in creating PS Private dialog
  • The instructions out of date flag was only getting reset if the user made a change in the edit instrs dialog. If s/he just pressed [OK] it would remain on.
  • Make libintl be another library we'll try to dlopen. This is complicated because on linux libintl is part of libc so there we must link directly. Further complicated because fink install script has a bug and doesn't put /sw/lib in DYLD_LIBRARY_PATH so we have to look there ourselves.
  • Locale strings aren't required to contain an encoding (which the locale docs incorrectly call a charset). Michal Nowakowski points me to nl_langinfo which will give me the encoding even if not specified in the locale.
  • Add pdf files to the standard format list for openfontdlg
• 14-Oct-2006
  • New translations: Greek, Simplified Chinese, German, Polish.
  • Using the help command would generally crash fontforge on cygwin.
  • Adobe uses the American spelling of diaeresis (that is dieresis), but some fonts contain glyph names with the European spelling. So add that as an alternate.
  • If a bdf font contained an unencoded .notdef glyph, then we crashed.
  • Adding Greek translations for font styles. (Thanks to Apostolos)
  • If we aren't dealing with a TeX math font, then the "Sub/super"script positioning fields of Glyph Info->TeX aren't meaningful. So grey them out.
  • Remove internal/external did the wrong thing in expand stroke if we also removed overlap and had counter-clockwise splines.
  • Alexej Kryukov points out that reverting a glyph with anchors lost the anchors.
  • Add simplified Chinese style names to our built-in lists.
  • Problem reading autotraced input.
  • The encoding change of last summer broke tfm output with glyph lists or extension characters.
  • If add extrema added a point too close to a start/end point expand stroke could generate NaNs.
  • FF could crash when adding extrema to an open contour.
  • Add Polish styles to the translation list.
  • Add a scripting function to set the gasp table.
  • Add support for the 'gasp' table.
  • Alexej Kryukov wants format 6 cmap tables.
  • After editing instructions the debugger would incorrectly complain that instructions were out of date.
  • Michal Nowakowski complains that "Instructions are out of date" warning appears too many times when doing a global transformation. Make it only pop up as a dlg once and thereafter occur as a LogError.
  • If a font contained only space glyphs it could still be marked onlybitmaps and not allow outline generation.
  • Show ATT did not translate scripts & languages.
  • Parsing an empty list of ttf instructions crashed Michael Zedler's machine.
  • Make type1 generation use (essentially) the same subroutine mechanism as type2.
  • Local subroutines were being freed twice in cff cid-keyed font generation.
  • Removing an alternate unicode encoding (Detach Glyphs) was broken in some cases.
  • FF failed to perform some simplifications on the last spline of a closed contour.
  • Various fixes for Michal N.
  • Add scripting interfaces to many of the ctype functions.
  • Add a scripting interface to the 'size' feature, now that Adobe has finally decided what to do about 'size'.
  • Had the wrong title for kerning state machine dlgs.
  • Redo the preference dlg.
  • The two preference items, AlwaysGenApple, and AlwaysGenOpenType, were a bad choice on my part and have lead to considerable confusion. I mentioned some time ago that I would remove them if there were no objection.

    There was no objection.

    I have now removed them.

    I also made the SetPrefs scripting function sort of work if the user tries to set one of these (it will set the appropriate flags in the default option set used by generate).

    The defaults of these flags were different on the Mac from everywhere else. This meant that a script could do different things on the Mac and on linux. That seems like a bad idea. I have changed things so that the default options setting for generate is always OpenType.

    THIS IS AN INCOMPATIBLE CHANGE AND SCRIPTS ON THE MAC COULD NOW BEHAVE DIFFERENTLY.

  • Patch from Michal Nowakowski to have ff number points after conversion to quadratic.
  • Try to provide disambiguation for all strings "New" so that in languages where "New" must agree in gender with the thing being created translators will be happier.
  • Remove the po/Makefile.in and redo the configure script to build it (so that I don't have to modify it each time I get a new language).
  • Extend the knife tool so that if Meta (Alt) is held down it will move any spline it intersects (rather than cutting the spline).
  • FF would crash if $HOME (the directory that is, not the environment variable) did not exist.
  • I guess it makes sense that if ff read in a compressed sfd file then using a simple File->Save command (not Save As) should produce another compressed sfd file.
  • When scaling a reference in otf output we had problems if the reference had more than one contour.
  • If we deleted many glyphs from a font we are sometimes left with kern pairs pointing to things not worth outputting. Make sure we don't generate kerning info for them.
  • Ligature carets were not being scaled by scale to em.
  • Patch from mfabian at suse.
  • If a user has a font with a glyph that is bigger than 16*em-size (in many dimensions) and that user generates a tfm file, then warn him that the glyph can't be expressed in a tfm file and truncate the size to the biggest legal value.
  • Attempt to mitigate the effects of X claiming to have a font when it doesn't.
  • Saving CID keyed fonts would look at garbage memory for its encoding data if not saved from the biggest subfont.
  • Add support for version 4 of the OS/2 table.
  • If Interpretation was set in a font then when generating a 8/16 type cmap fontforge always went down the traditional chinese (big5) branch.
  • MS has changed their indic shaping engine. It still uses the same features but they are applied differently. To disambiguate this change they have created a parallel set of script tags for the indic writing system. FontForge should recognize these and use them when appropriate.
  • When Adobe first released fonts with a 'size' feature they did not follow the opentype specification. In August of 2006 they announced an incompatible change to bring fonts produced after that date in line with the specification. Unfortunately there are programs which depend on the old, broken format.

    Legacy programs will not work when given a font in the correct format. New programs should be able to handle either correct or broken 'size' features.

    A few of Adobe's programs make use of the 'size' feature, and XeTeX does. I know of no others (besides font editors). Both intend to upgrade their programs... So this may be less of an issue than I initially feared.

    FontForge should now be able to read in a size feature in either format and gives the user the option of producing 'size' features in either format on output.

  • For Japanese fonts we tried to use cp932 (a variant of sjis) but then halfway through didn't know what to do with it.
  • Another fix for saving sub-opentype fonts from a cid font.
  • From a script, if a user asks to generate a ttf outline font with ttf bitmaps from a database with no outlines then technically this isn't supposed to work. But it used to and it's easy enough to tweak so that it works again.
  • Update to Unicode 5.0
  • Again when adding accents to the left of a glyph we should shift the glyph over to give it a reasonable lbearing.
  • When building an accented glyph, we generally want use_my_metrics set on the reference. But if we add an accent to right (horn) or left (psili/varia on cap greeks) we want to turn it off. We did turn it off on the right side, but not on the left.
  • Philipp Poll points out that ScaleToEm doesn't scale the OS/2 metrics (and some other stuff). Actually it does, but font info did the scaling BEFORE setting the values. So the metrics get scaled, and then reset to the values in fontinfo (which usually means they don't get changed).
  • Fix some memory leaks.
  • Some fixes for pcf loading.
  • Various fixes to CID keyed fonts.
  • Even the pretty ttf approximation routines need to worry about diametrically opposed control points.
  • When generating multiple type1 fonts, output the bounding box of each as that of the full font. This will make it vaguely likely that applications will get the same line spacing.
  • Oops. The vertical metrics bit wasn't preserved when flattening a cid font.
  • Ack! Window titles and such were being interpreted in Latin1 and not in UTF-8.
  • Interpretation of PostScript arc operator failed if the two angles were 360 degrees apart. Normalization made them both the same.
  • Try to clean up bitmap only sfnt generation.
  • pcf wasn't in the Open Font dialog format list (nor was pmf, nor pdf)
    Add a filter for TeX bitmap fonts.
  • Provide a mechanism (if a preference item is set) for picking a cmap when loading a font.
• 22-Aug-2006
  • Add a pane to the Preference dlg allowing user to control some of fontforge's hint creation.
  • Someone complained that Find Problems->Path Direction got the wrong answer. It got the wrong answer because they had a self-intersecting path (which screws up all concepts of in/out-side. So try to make it clear that they should do a find intersecting paths first.
  • Update my opentype scripts list (in particular include music).
  • Due to a typo, ff thought pretty much everything outside unicode BMP was in the hebrew script.
  • Add a new command to the point menu which allows you to specify a spline and one coordinate (either X or Y) on that spline. FF will calculate the other coordinate and add a point there.
  • CODE2000 contains an anchor class with only base glyphs (no mark glyphs) as part of a set of anchor classes. FF crashed.
  • Patch by Nix. svg.c did not compile when designed to be linked directly with libxml (instead of dlopening it).
  • Redo the font open dlg so that there's a list of standard filters and a mechanism for changing that list. Remember the last filter used.
  • If a popup menu is huge it used to appear under the cursor. This meant that even if the user didn't mean to s/he would have selected something and press/release would not be a no-op. This was a problem with the Font Info ->TTF Names
  • Patch from SUSE. Use of uninitialized local variable. But the result is never used, so just remove the offending lines.
  • Emulate gtk's boxes and use that to do automatic dialog layout. Get the font info dlg working (and then several others too). This means that the layout is no longer optimized for English, but will look better in other languages (I hope).
  • Add an information to Font Info to display what unicode character ranges are supplied by the font.
  • Philipp Poll points out that when generating versioned fonts (ie. Ambrosia-1.0 rather than just Ambrosia) you don't want the filename to default to the fontname (the fontname will be Ambrosia), instead you want some way of specifying a default filename. So add that to the Names pane of the Font Info dlg.
  • Try to keep track of font creation and modification times. NOTE: This is different from file times (which is what appears to be stored in the 'head' table of an sfnt file). Add a new table 'FFTM' to our sfnt output with these timestamps (and the timestamp of fontforge itself, just in case we ever care).
  • Mark classes could not be set from the selection.
  • Ben Laenen points out that if an order2 glyph has a dependent which refers to it using point matching and we keep changing the point number on the glyph then we'll keep getting warning messages. He suggests one message.
  • Empty glyphs weren't marked with red Xs any more.
  • In font compare, when comparing two contextual/chaining lookups, I forgot that class[0] was always NULL.
  • Open contours caused the ttf point renumberer to crash.
  • A font can have many kerning pairs. More than will fit in one GPOS subtable (which essentially has a 64k data limit or a bit more than 10,000 kern pairs). So be prepared to generate many sub-tables each dealing with a subset of the (first) glyphs.
  • If an apple mutually exclusive feature had a default setting which wasn't 0 then my previous patch didn't work.
  • If guessing bluevalues found no OtherBlues, fontinfo->PS Private would create an incorrect OtherBlues value [0]
  • Support clipboards containing image/svg
  • Ralf Stubner points out that the maximum zone size of bluezones is determined by 1/BlueScale. So adjust BlueScale if the max zone is too big.
  • Merge fonts didn't merge anchors and got into a terrible mess. Let's merge kern classes and state machines too. Merging kern classes may be a bad idea. We don't really merge them we just copy them so that if the same glyph is mentioned in two classes unexpected things might happen.
  • ff could crash on an sfd file containing erroneous anchors (such as might be generated by the merge fonts command).
  • Once upon a time, Apple was so concerned about accuracy of floating point operations that they required all operations be performed in IEEE extended (80bit). This had the interesting side effect that a program which conformed to Apple's standards could not also conform to ANSI C (which required operations use double temporaries for double values). I find it somewhat ironic, therefore, that gcc on the mac doesn't even promote floats to doubles. FF needs the extra accuracy provided by a double (an extended type doesn't seem to provide additional assistance unfortunately). So on the mac we got extrema that weren't, which meant that remove overlap didn't.
  • If we had a mac feature which had multiple subtables (ie. a state machine for a contextual substitution, and a non-contextual substitution too) then we produced a bad chain in the morx table.
  • Merge & Simplify would sometimes generate solutions using control points which were diametrically opposed to what they should be.
  • Yet another attempt to do a better job of autohinting.
  • Add a "Revert to Backup" command.
  • Put code in to correct for single bit (sometimes double bit) errors when search from extrema. This is at a lower level than the previous patch.
  • Output of extension lookups wasn't working
  • Fixed a problem in the ttf names pane of Font Info.
  • A slight improvement to magnifying by dragging a rectangle.
  • The first time you click on a transition in a Font Info->Mac SM->Indic state machine, ff would crash.
  • Tihkomirov points out that it isn't clear how to delete a ttf name from the name pane of fontinfo. So add some popup help. The language & string ID menus in that pane were sorted on the wrong key
  • Was not preserving (unparsed) tables properly. Several bugs.
  • When loading an apple bitmap font (with no hmtx tables) we need to dummy up widths for the outline glyphs.
• 15-Jul-2006
  • Apple's bitmap only fonts should not have an 'hhea' nor an 'hmtx' table
    Dummy up outline glyph widths when we don't have a 'hmtx' table by looking at bitmap widths for that glyph.
    ftxvalidate complains if the vertical line metrics are 0. Even in a strike with no vertical metrics information. So dummy something up.
  • Was getting the flags wrong for turning off features in an apple morx chain.
  • When outputing mark-to-base sub-tables with multiple marks I neglected to output an NUL offsets for the anchors that weren't there. So if a glyph had only one anchor (instead of two, or however many) then the offsets would be all wrong.
  • The SetPref() scripting function was misdocumented to be called SetPrefs() so accept both names.
  • When building an accented letter out of glyphs with an anchor class ff would increase the advance width of the accented glyph by the size (xmax-xmin) of the accent.
  • Expand the functionality of the ruler tool so that it also shows curvature.
  • When importing a bitmap font from a mac binary or an sfnt and the encodings of the destination and the bitmap did not match, then things got rather screwed up and might crash.
  • patch from KANOU. the ref check routine in glyphcomp compared one reference to itself.
• 03-Jul-2006
  • Type2 charstrings have no closepath operator. I used to think that meant I could not omit the final vector. In fact it means I can omit the final vector AND the closepath operator.
  • If we had a font with no complete code pages we used to generate a version 0 OS/2 table without codepage info. But now windows refuses to install such a font. Instead generate a bogus codepage field which claims to have latin1 (even though it doesn't).

    Windows won't use a font where the codepages field is 0.

  • When generating a FOND for a bitmap only sfnt, we created dummy 'NFNT' resources with the wrong resource ids. Which meant we got garbage when the font was used.
  • Make the TabSets in the font & glyph info dialogs run vertically rather than hor izontally. (Add an option to get the old display back).
  • Approximate Spline from Point Slopes sometimes set the control point while leavi ng the nonextcp flag on.
  • Add Extrema only looked at how close a point was to the extrema in one dimension. While that dimension is more important, make a less stringent check on the other.
  • Expand Stroke removes "U-turns" from splines because they confuse it. But the routine that did that got a bit rambunctious and did the wrong thing on a circular arc.
  • If a font did not contain an OS/2 table then we didn't retain the 'hhea' ascent/ descent.
  • Various fixes from Matthias Kilian of OpenBSD.
  • When asking for default 'subs' features, every glyph gets assigned U+2080.
  • Display the curvature information at each point. In Point Info dlg, and in a pop up window when editing cps.
  • Add two new menu commands/scripting commands:
    • Canonical Start Points (CanonicalStart())
    • Canonical Contours (CanonicalContours())

    The first sets the start point of each contour to the leftmost point, while the second orders the contours by their leftmost points. This will slightly reduce the size of output charstrings. It will also make more things available to be put in subroutines.

  • Print Multi-Size & Full Page printed the wrong glyphs from the font view. Broken since the encoding change.
  • When stroking a contour that makes a 180degree turn, ff did not give it a semi-circle if linejoin was set to round.
  • SPLCopyTranslatedHintMasks didn't work if there were more than one contour being copied.
  • Nathan Summers points out that if the XUID is garbage the ff can go into an infinite loop when generating an otf/cff font.
  • Somehow the XUID and UniqueID fields crept off the bottom of the fontinfo Names p ane and weren't visible or modifyable. Give them their own pane now.
  • Apply some patches (with some changes) from Nathan Summers to the truetype debugger.
  • The property count in BDF fonts is wrong.
  • Ralf Stubner points out that the notdef glyph in an otf font should look like that in a ttf font, and not like that in a type1 font. (that is there should be something in the glyph).
  • Add support for group "encodings" which have empty spaces for unused glyphs.
  • make another attempt to find a good default sli for glyphs with no default script.
  • Typo in scripting function DefaultATT.
  • The View->Display Substitutions has been broken since the encoding change last summer.
  • Mark glyphs with out of date instructions differently from those with no instructions.
  • The last change to the cvt table generally did not get made.
  • Make glyph compare check for point matching in anchor points and in references.
  • Add a warning message about out of date instructions when generating a font.
  • When generating a font with anchor points matched to ttf points, ff failed to allocate enough space for them and offsets got all confused.
  • Oops. Some variables which should have been static were dynamic. Which meant we kept dlopening libfreetype.
  • If we had an anchor point which positioned itself with point matching, and we ge nerated an otf font then we'd output a lookup which tried to do point matching i n postscript. Only do point matching if we are sure we're in a tt font.
  • Add a preference item to control clearing of tt instructions.
  • Printing a multi-size page of a truetype font crashed.
  • Anchor Point dlg didn't work well with matching tt points.
  • Add a field to allow users to search for a glyph name within a kerning class.
  • Interesting. In kerning by classes, if the coverage table does not match the glyphs specified in all the classes (for the first (left) glyph), then any glyphs in the coverage table but not in any class should be treated as being in class 0.
  • Add something to the font view's copy from menu to control whether truetype instructions are copied or not.
  • Add some more problems to find problems
    • Check for glyphs with both contours and references.
    • Find out of date point matched references.
    • Find references with transformation matrices which can't be expressed (as refs) in truetype
    • Find references with transformation matrices which aren't translations (and so can't be expressed (as refs) in type1/2 fonts.
  • Allow the user to turn off composite info in afm files.
  • otf kerning classes were dumped in the wrong order.
  • In GaramondPremrPro.otf (version 001.000) Adobe has coverage tables (for some context chaining subs) which have the glyphs mentioned multiple times. I suppose this is legal, but it is annoying. I want my strings to have each glyph appear at most once. So add a little thingy to remove duplicate glyphs when we generate a list of glyph names.
  • Adobe has depreciated the use of PUA to represent small caps, old style numerals, etc. Add a preference item so that ff will no longer recognize "a.sc" as having a PUA encoding.
  • The tool tip for the preference item "NameListForNewFonts" was wrong.
  • Horizontal scrolling in the show att dlg was too slow. It was by pixel. Make it be by something which approximates the width of a character.
  • Add a new scripting command to give a font's bounding box.
  • Sigh. Within a GPOS lookup, sub-tables are not additive. Only one will be applied at a time. So order the 'kern' lookup so that pair-wise kerning comes out first (to override class kerning).
  • Was sometimes outputting mkmk base-mark attachement points as the mark attachment point.
  • Clytie Siddall has many suggestions about how to make the translation process easier for translators. He also tells me how to insert contextual comments into the program source, and provides many fixes for typoes.
  • The OS/2 version 1,2 has a different defn of avgCharWidth than OS/2 version 3(,4) does. Display of CJK fonts is supposed to break under windows if using the v3 defn.

    Use the appropriate defn for the version we happen to be outputting.

  • Autohinting crashed on glyphs in an order2 font if the glyph contained both an outline and a reference.
  • Pressing the help key in the main windows went to the overview page, not to the contextual help for the window.
  • Add a warning when 'CFF ' widths and 'hmtx' widths differ (when loading a font).
  • Set the default and normative widths too late after the cff subroutine change.
  • Make double clicking in the list of kern pair/anchor combinations bring up either Kern Pair Closeup dlg or Anchors Away.
  • The Anchor control dlg didn't work after the encoding change.
  • Add a popup message to explain what "Cor:" means.
  • Doing a Close() from a script didn't work if the windowing system were open.
  • Yet another bug in type1 hinting output.
  • Some versions of realloc don't like allocating 0 bytes.
  • Users expect ff to recognize a glyph name "Diaeresis", even though Adobe (following standard American, but not European spelling) spells it "Dieresis".
  • GhostScript has canniptions when it gets a version 3 OS/2 table in a ttf file.
  • Can't add a flex hint when there's a hintmask at the mid point.
  • The hintmask part of PointInfo would not let you clear a hintmask.
  • Add a new command to the fontview: Edit->Select->Select by Name
    and a new scripting command: SelectAllInstancesOf

    Both will select all encoding slots containing the named glyph (or unicode value)

  • Add scripting command: RemoveDetachedGlyphs
  • Was setting typo linegap to hhead linegap.
  • Using the X clipboard to paste into the character view has been broken at least since I added multi-layered editing.
  • Add a pref to control whether we prefer to use spacing or combining accents when building accented glyphs.
  • Spacing in the group dlg was sometimes off.
  • More hinting fixes. If the edges of a stem aren't quite parallel we had problems
  • Add some new built in variables: $order, $em, $ascent, $descent
  • The 'TeX ' table was output incorrectly if glyph's order wasn't the same in the sf as in the ttf.
  • Allow the [*] Back visibility checkbox to control whether the grid fit lines are visible or not.
  • If scripting is in verbose mode, flush the output buffer before calling a built in procedure (so any errors it generates will at least show the right line).
  • Ambrose LI points out that the master font in a cid keyed font should match the ascent/descent of at least one of its subfonts (so that the typoAscent/Descent f ields will default correctly).
  • Spend some quality time with valgrind looking (oh, and fixing) memory leaks.
  • Make ControlAfmLigatureOutput (scripting command) apply to ofm/tfm files too.
  • Oops. When parsing mac state machines I was using sizeof(long) when I should have said sizeof(int32). It gave rather bad results on 64bit machines.
  • Add a scripting function to load a user defined cidmap file.
  • Hmm. sizeof(point)==8 on AMD64, but structures are aligned to 4. This means our chunk alloc routines complain. So if we're given something that isn't in units of sizeof(pointer) then just round it up.
  • lookup ordering within GSUB has been totally broken at least since I added the code to coalesce lookups for kerning.
  • Kanou suggests changing function name that includes Char to be Glyph. Well a few do really mean Char, but the rest should be Glyph. Leave the old names there bu t document them as deprecated.
  • Make CID keyed fonts use the new subroutine mechanism too. We can use global subrs now too!
  • Oops. Familyname can be NULL, and we were indescriminately checking to make sure its length was less than 31.
  • Make a better use of subroutines when generating otf fonts. Doesn't work for CID fonts yet.
  • We crashed if we changed from one cid sub-font to another when compacted.
  • Oops. I made ff work with ft22 if the debugger was being built... but it failed to compile if people didn't build the debugger.
  • Oops a scripting function was marked as not needing a font when it did (HasPreservedTable & friends).
  • More for font compare. If there is a difference in a glyph in the two fonts add the option for including the outlines of the second glyph into the first. If the first font is missing a glyph, add an option to insert it with the outlines in the background.
  • Good heavens, someone wants to play with acorn2sfd. Make it compile again.
• 08-Apr-2006
  • Patch by Mike Frysinger from gentoo. Use the HOME environment variable in preference to /etc/passwd.
  • Werner wants to see raw data in the points window of the debugger. My guess is that most people won't be interested in this mode (and it makes the window a bit uglier) so I've made it configurable. --with-debug-raw-points
  • Write sfddiff as a fontforge script and put it back into the install process.
  • Add code to font compare so it can do the old sfddiff --merge option (if two corresponding glyphs differ, add the outlines of the second to the background of the first).
  • Someone wanted acorn2sfd! Make it compile again.
  • Make configure look for libgif
• 06-Apr-2006
  • Patch by Ian Jackson of Ubuntu. When in an empty glyph (one not WorthOutputting) then doing a View->Show Grid Fit would cause a crash.
  • The PLRM says that formfeed is whitespace. The T1_Spec says that it is not. After an eexec the T1_Spec is right (and ff was following the PLRM definition).
    Also add a check to insure that the first character we generate in pfb encryption isn't a white space character.
  • On some windows systems (mine at least), windows will refuse to install a postscript font if the fontname or familyname of that font is longer than 31 characters. These are the names in the Font Dictionary of the postscript font itself, not the names in the 'name' table (for otf fonts).
  • Make ff compile with freetype 2.2.0
  • The popup menu of text fields was garbage.
  • Make a change to pfm output which might allow non windows ANSI encoded fonts to work.
  • Allow the freehand tool to work in quadratic fonts.
  • Didn't always refigure splines properly when using Get Info in quadratic mode. Patch by Michal Nowakowski.
  • Trying to invoke the KDE browser (for help) didn't work. Patch by Dimitar Zhekov
  • Still losing instructions if contours begin with off-curve points.
  • if a font had a defined notdef glyph, then FF would usually enter it incorrectly in the 'hmtx' table.
  • In the Select menu, Width had an extra underscore.
  • Wasn't looking in the right place to get the right weight name in an otf font.
  • People have been complaining about the UI for drawing quadratic (truetype) curves. It was consistent with the UI for drawing cubic (postscript) curves, but it seems hard to draw quadratics that way. Rather than concentrating on the on-curve points (as in cubics), now concentrate on the off-curve points. So the pen-tool behaves quite differently now in quadratic mode. As does Element->Get info.
  • A patch added back in 2004 to make vertical substitutions work for Japanese (to work around a bug in Windows to be precise) had a bug itself if more than one language was involved.
  • When outputting an inline copy of a reference in otf fonts with translated hintmasks we did the wrong translation for vertical stems and got garbage results.
  • New command to allow user to enter X,Y coordinates of a point and have that point be selected. I want this for debugging. It might be useful generally.
  • Fixed problems with type2 subroutines containing glyphs with conflicts.
  • ReplaceWithReference had too large a fudge factor and found macron when it should have found macron.sc in minionpro-regular.
  • If an alternate/multiple subs had more than 30 components then we lost all those after 30 when generating opentype GSUB table.
  • Hmm. Sometimes least squares does give the answer we want for merge. So give it a try.
  • Patch by kanou to fix some translation problem.
  • In the shape tools (round rect, ellipse): Instead of generating a quadratic shape at the start, generate a cubic shape and convert to quadratic. The convert routine now knows about how to find integer control points and such.
  • Patch from Kanou to get more grid size values for looking at bitmaps.
  • If a glyph starts with a control point then ff would sometimes throw out the ttf instructions, thinking it wasn't numbered properly.
  • A merge in a quadratic font produced a straight line. Even when it shouldn't have.
  • Change namelists to support non-ASCII glyph names.
  • Oops. Some hinting scripting commands had a "b" prepended to their name by mistake. Sigh. From now on support both names.
  • When pasting a glyph with hints into an empty glyph, we shouldn't set the "in need of autohinting" flag. (Actually, we might should if the original copy were done in a glyph with out of date hints.... but that gets more complicated).
  • Patch to make loading bdf fonts more efficient.
  • Add Kanou's "fontview" font to the list of fonts searched by default for the fontview window.
  • FF did not mark ghost hints as different from normal hints.
  • Interpolating fonts broke when I did the encoding change last summer.
  • Patch from Kanou to prevent the hash function from returning negative numbers. (and crashing)
  • Add a menu and a scripting command to compare two fonts.
  • Oops. There's a stray underscore in a field in font info.
  • Refresh font view(s) after changing glyph name/unicode in glyph info in case those are being displayed.
  • If a name entered in the goto dlg was a single unicode char, then goto would treat it as such. Normally this is fine, but a user might use the underscore character (for example) as a glyph name and not assign it to U+005F. So check for the name too.
  • Rewrite the cubic->quadratic conversion code. It now looks much prettier and tries to position intermediate points on integer (or half integer) locations so there won't be unexpected rounding later. It's slightly less accurate though.
  • Many changes to the postscript hinting code.
  • If a glyph contained references but no alphabetic characters then Glyph Info would crash.
  • Scripting command AskUser doesn't work unless FF_SCRIPT_IN_LATIN1 is set. Patch by Kanou.
  • In some error messages I ended a sentence with a comma. Make it a full stop.
  • Patches from Kanou to make Japanese translation better.
  • Kanou points out that the code which handles modifications to the unicode char textfield (in glyph info) complains about the wrong things and he provides a patch.
  • Hand tool failed to scroll background images (probably failed on hints too).
  • When doing a copy splines with hintmask operation we neglected to refigure the spline after transforming the end points. This was ok since we never used that info -- except if the splines were quadratic and we needed to convert to cubic. So when generating a ps font from a quadratic database, and the font contained references with scaling/rotation/etc. then we'd get erroneous results.
  • In Find Problems, the Flipped Reference check could fail if the reference contained a contours which intersected (themselves or each other).
  • Make ff aware of ExtraLight as a valid weight when setting a default value for the PS weight string.
  • When the user clicked on a textfield with attached pulldown list, ff would get an event saying the first entry on the list had been chosen. This meant that the Tag dialog (used for new ligatures/substitutions/etc. would show the first tag in the list instead of blank.
  • Include some glyph name synonems from Richard Kinch's TrueTeX list.
  • Determination of whether a ghost hint is top (-20) or bottom (-21) should depend on the current contour, not on the glyph as a whole.
  • If a glyph consists solely of references then add _<ref-name-1>_<ref-name-2>...<ref-name-n> to the list of possible names. Sometimes also add uniXXXXXXXX.
  • Kanou wants to translate some more strings.
  • Barry.SCHWARTZ wants to be able to reencode a font to compact from a script.
  • Barry.SCHWARTZ points out that my "cfg" files contain no newlines. Shows how much use they've received:-)
  • Barry.SCHWARTZ points out that when generating an afm file for an otf font only the first 256 characters were given encodings.
  • Simplify much improved.
  • Make simplify for quadratic splines aware of the peculiarities of truetype. So one simplification is to make a point interpolateable. And removing a point is not always good if it means points that were interpolateable become non-interpolateable.
  • After examining (and stealing with permission) Huw Davies's code in wine/tools/fnt2fon.c I believe I can now produce .fon files.
  • Add support for reading bitmap references (from an sfnt bitmap)
  • Patch by Kanou. cvknife didn't compile in one mode.
  • Typo in an error message
  • Add a command to show points of inflection (also change View->Mark Extrema to show where extrema occur on a spline if not at the endpoints)
  • Patch by Michal Nowakowski. Uninitialized variable in simplify dlg.
  • glyph names for unencoded glyphs were always "uniFFFFFFFF" should be "NameMe-<n> broken by last summer's encoding changes.
  • Werner wants CC (composite character) entries in afm for GPOS mark-to-base features.
  • The switch to utf8 last summer broke font info->Size
  • Allow users to ask for both 'GPOS' and old-style 'kern' tables. FontLab does this. The OpenType list is generally disapproving of the idea.
  • If we tried to put more than about 10000 kerning pairs into one old-style kerning sub-table then we'd overflow various shorts. So split it up after about 10000 kp into several different sub-tables.
  • Round to Int screwed up interpolated points.
  • When generating a tfm file don't generate a width bigger than 16*em-size (even if the font says to) because TeX will throw it out.
  • Alexey Kryukov's patch to fix an error message.
  • Printing a CID keyed truetype font to a pdf file produced a very odd ordering of glyphs (compacting some and omitting others and getting errors when it ran out of glyphs but asked for more).
  • Alexey Kryukov suggests repunctuating an error message.
  • Johan Winge suggests an improvement to converting cubic splines to quadratic: If we end up using a line when looking at the spline as a whole, try breaking it at the point(s) of inflection first. It seems like a good idea.
  • Johan Winge also points out that I identify points as tangents when they should not be. This is because I use one measure of sloppiness no matter how close the cps are to the base point when determining colinear. It's probably best to have a variable measure depending on how far the cp is from the base point.
  • Alexey Kryukov thinks Add Unencoded Slots should work in the compacted view by making the thing have a custom encoding.
  • Scripting reencode didn't get rid of the "compacted" mark.
  • Ralf Stubner points out that I produce type1 charstrings which end in "seac endchar" when a simple "seac" is sufficient. He's right.
  • When generating opentype tables in from a single sub-font of a cid-keyed font (ie. user asked for a ttf font generated from one sub-font) ff would crash.
  • We didn't hint bold fonts well (anything where stems were bigger than em-size/10)
  • When generating a type1 font with a glyph containing a reference that had to be unlinked, then the vertical stems in that reference would not be hinted.
  • When guessing points at which to attach hints in type1/2 read in we failed if the points were too widely spaced. Be a bit more lenient.
  • SetCharName didn't check its flag value properly.
  • Uncompacting a font caused memory problems.
  • Fix simple crash in metafont.
  • When importing unencoded glyphs into an already existing font, we would sometimes fail to find the matching glyph (and would create a new one).
  • If a font contain an encoded glyph called ".notdef" and that glyph was not the first glyph, then type1 fonts had a .notdef glyph at a random place in the font. Now freetype seems to renumber glyphs so .notdef is always glyph 0 which fontforge did not expect and produced some strange results when rasterizing (tested on 2.1.10). Fix is simple: always make notdef be first in the chars dictionary and then all is happy.
  • Add new scripting command to compare two glyphs. I want this for testsuites, other people might find a use for it.
  • gniibe reports that the scripting function Export() produces bitmap images with the wrong format and provides a patch.
• 9-Feb-2006
  • Instead of refusing to do expand stroke/remove overlap for order2 fonts, why not convert the contours to order3, do the function and convert back.
  • Once more we are numbering points badly.
  • If we clear a glyph which is being referred to (and we don't unlink it first) then when we save it into an sfd file we get a reference to a glyph at orig_pos -1 (because the cleared glyph isn't stored in the sfd) which causes an error when reading it back in.
  • If a copyright message were an exact multiple of 100 characters long, fontforge would screw up memory (and potentially crash) when reading it in from an sfd file.
  • The Merge/Simplify command (which Add Extrema now calls) had problems when the control points were very close to the base point so that the point was (to the human eye) a corner point.
  • Add support for the r2l script N'Ko. MS says opentype will use script 'nko ', ISO says the tag should be 'nkoo'. Follow opentype.
  • Kanou tells me that Syriac & Thaana are also r2l scripts and should be added to my list.
  • If FF was given multiple fonts on the command line the sizes of the font windows would jump around in a very peculiar fashion.
  • Add the concept of a namelist. Users can specify what names they want for new glyphs, or can force an entire font into a new naming scheme. Changes to
    • FontInfo
    • Open
    • Generate (Family)
    • Preferences
    • and the Encoding menu.
    • New scripting functions:
      • LoadNamelist(filename)
      • LoadNamelistDir([dir])
      • RenameGlyphs(namelist)
      • Extended Generate (scripting) to take a namelist argument.
  • If users didn't have libuninameslist installed we were still generating spurious ligatures for numero and TM. (and others)
  • KANOU points out that if you comment out a bdf glyph (with COMMENT) fontforge sort of reads it anyway.
  • Kanou needs to distinquish between the script latin and the language.
  • SplineRemoveAnnoyingExtrema1 got tweaked a little too much in attempting not to generate insane control points.
• 25-Jan-2006
  • If you try to create a bitmap only sfnt but provide pixelsizes for no valid bitmap strikes, the fontforge would crash.
  • Chia points out that I should use the PIXEL_SIZE property to set a bdf font's pixelsize (instead of using font ascent+descent).
  • Trying to rasterize a glyph with a coordinate of 1e7 caused a crash.
  • The change to make Add Extrema remove points near the extrema failed to consider the possbility that some of those points might be corner.
  • In Gentium Roman, when searching for bluevalues, the baseline is further than 1 standard deviation from the mean of glyph ymin. Change the range searched so that 0 is always included.
  • In Font Info->PS Private, if we selected OtherBlues and pressed [Guess] it did not change the OtherBlues value.
  • The gettext patch changed Histograms->Blue Values menu item into a line.
  • When changing the order of a font, ff would ask if you minded losing the undoes. There's no point in asking the question if the font hasn't been modified (it has no undoes).
  • The Point Info dlg should enforce that curved points cps are colinear (so if the user changes one cp, the dlg should change the other to be in the opposite direction.
  • There was a path though PointInfo that called SplineRefigure3, which meant that splines got marked as cubics even when they should have been quadratics.
  • Using a rotated elliptical pen in expand stroke did not work
  • Add a homedir button to the file chooser dialogs.
• 17-Jan-2006
  • Make the Import() scripting command accept {bdf,pcf}.gz files
  • Reading utf7 (ttf name) strings from an sfd file was broken if one of the characters had bit 15 (high bit) set and its position in the string mod 3 was 1.
  • My horizontal line metrics in the sfnt bitmap strike header were incorrect
  • Use the horizontal line metrics (if they work) to set the ascent/descent of a bitmap strike.
• 15-Jan-2006
  • The change to fold all COMMENTS into one bdf property broke the sfnt 'BDF ' table routines.
• 14-Jan-2006
  • Change AddExtrema so that if an extremum is very close to an old point it will add the extremum and then remove the old point.
  • Redid how bdf properties were handled
    • I had the definitions of some properties slightly wrong
    • I was generating some obsolete properties
    • When I generated CHARSET_ENCODING for iso8859 encodings I had an extraneous "-" in the value.
    • retain the properties when loading a bdf, pcf font
    • Add a dialog which gives user control over bdf properties
    • design a 'BDF ' table for SFNT files which can contain bdf properties and make FontForge read and generate this table.
  • Ted Packard points out that I output "big metrics" in sfnts inconsistently. This patch should make them consistent and right.
  • Restore "compact" views to their old glory. Ie. they can now be uncompacted (and uncompaction happens automagically when generating a font).
  • When reading a GenTags: line from an sfd file stored with CRLF line terminators, the next line would be lost.
  • ff couldn't handle xrefs which were broken in bits in pdf files?
  • When reading a ttf font from a pdf file we found no encoding info. This caused us to crash because various routines expected an encoding map. Add a dummy (Original) map if not exists.
  • Wasn't numbering points in references properly.
  • Gleep! Ever since the gettext code went in, the Generate fonts dlg has refused to generate "No Outline Font" and "No Bitmap Font".
  • Herbert Duerr from OpenOffice tells me that their OpenSymbol.ttf font when produced by fontforge does not work on Win98. This is because Win98 doesn't like the codepages bits ff produces and refuses to use any glyphs in the font. He submitted a patch to make ff produce a version 0 OS/2 table (with no codepage info) for this font. Win98 will accept it then. His patch was against an earlier version of ff and needed to be modified for the current version.
  • Patch by Raph Levien which finally shows me how to use FT_Outline_Get_Bitmap properly.
  • If the user has never called PrintSetup (or if the user has not called PrintSetup this session and has not loaded prefs), then fontforge will invode setpagedevice on a pagesize which will fit on either A4 or US-Letter. The PS manual says a printer should refuse such a size unless it has paper within 5 points of it. So on printers which follow the spec (not mine) nothing will print as it waits for paper.

    Now if the pagesize has not been set, don't specify a PageSize to setpagedevice.

  • The PrintSetup scripting command should not require a font.
  • Add a mechanism so that users can prevent ff from making a point implicit. Useful in that point needs to be instructed.
  • New copyright notices for 2006
  • Make the pen tool work better with quadratic splines (make it default to having control points which cause the point to be implicit).
  • Order2 elipses were not drawn accurately.
  • Point numbers sometimes didn't get properly renumbered after changes. If we have references & splines in a (truetype) glyph then show the numbers as "?".
  • When a charview displays numbered (truetype) points, then it shows all the control points. But the control points are not selectable. This is not a good thing, so make them selectable.
  • The font display to pdf did not include a title at the top of each page.
  • Peter Denisevich points out that when we print something we say "%%Pages atend" instead of "%%Pages (atend)". And the latter is correct.
  • Werner points out that my type1 (and type1 cid) fonts contain a %%DocumentSuppliedResource line, and they should not. The next level up should do that when it includes the font.
  • Show .xpm files as images we support in import image dlg. We've always supported them, might as well show them.
  • Check the glyph class when deciding what anchor to use.
  • Anshuman Pandey wants to be able to set the glyph's GDEF class. I've never understood why the default behavior wasn't good enough, but it's easy to do.
  • One of the fields in Glyph Info had some garbage initialization.
  • Add a default anchor point type to the AddAnchorPoint scripting command. FF will try to guess an appropriate value.
  • Add some standard window sizes to the fontview window. I'm tired of having the windows sized to other people's standards whenever I review a bug.
  • There was a path through expand stroke where a control point was marked as non-e xistant by mistake.
  • Expand Stroke would leave contours with the reversed orientation if it had to do a remove overlap internally.
  • Oops. The routine to save the encoding prefs file could not handle an encoding plug-in.
  • KANOU points out that the origin pulldown list in the transform dlg was not being translated.
  • In the preference dlg all of the [...] file browsers had the title "Call Script" which isn't appropriate for most of them.
  • Typing .* and pressing [Filter] did not show hidden files in the file chooser.
  • the code to check for intersections failed to check if there were a crossover in the last tiny section of a spline before its end.
  • Was generating ofm files when creating bdf fonts. Seems unnecessary.
  • Was producing a EBLC table with bad first/last values for first glyphs.
  • Hmm. Not all systems have tzset. Add a configuration option for it.
  • In scripting, a return within a loop sent ff into an infinite loop.
  • Kanou's patch for adding a new flag to scripting Import, to clear out the layer first.
  • Patch from KANOU to add SetGlyphChanged scripting function.
  • If an implied point became real, this fact was not reflected on the display until the user forced a point renumbering.
  • A mouse click in an empty metrics view references garbage memory -- And crashed Kanou's session.
  • Loading a large bitmap font left bits of the backmap of the encoding map uninitialized.
  • Loading a pcf font caused a crash.
  • Moving the control points of an implied point should move the implied point. And we should indicate which points are implied so this behavior doesn't surprise.
  • Michal Nowakowski points out that renumbering points in a glyph should turn off point matching in references.
  • Add support for bitmaps with vertical metrics.
  • Wasn't storing the vendor id in the OS/2 table properly.
  • Rounding error present on PPC which was absent on *86 machines.
  • Werner would like pixel size info to show on the title of a charview when debugging.
  • Added a new menu command "Point->Center Between Control Points" for truetype.
  • When manipulating menus with keys (rather than mouse) Werner found a crash.
• 5-Dec-2005
  • Creating the first bitmap view might cause a crash -- depending on how the linker organized memory.
  • Loading an sfd file now checks to see if there are any glyphs with splines in them, and if there are not turns on the onlybitmaps flag. Someone pointed out that if you inadvertantly edit in the outline view -- even if it is nothing significant, then the font is forever after marked as outline. This isn't a perfect fix, but it will prevent the worst from happening.
• 1-Dec-2005
  • Use GNU gettext rather than my nomen routines to handle translation of the UI..
  • Add support for plugins.
  • I added a check in parsing tfm files to make sure I didn't read outside the bounds of the kern table. But the table size was expressed in ints (32 bit units), and my index was in bytes, so I frequently exceded it.
  • If we used Get Info on a reference in a ttf font, and that reference did NOT do point matching we would generate an inappropriate error.
  • My support for using endchar as seac in type2 fonts only worked in bare cff fonts, it did not work if the cff were inside an sfnt wrapper (opentype).
  • I have decided that I will now store all postscript strings in utf8 (copyright, weight, familyname, etc.) They SHOULD all be ascii. But the occasional copyright mark would sneak in. We went into an infinite loop on one such because that was an illegal utf8 string. So fix a number of problems related to this.
  • Switch from using an internal routine to using freetype to rasterize b&w bitmaps when debugging ttf instructions.
  • Make the gridfit/debug settings sticky across invocations (store in prefs).
  • Add popup info showing the level of greyness in a truetype font being debugged in anti-alias mode.
  • Werner suggests altering the dynamic range of anti-aliased rasters in the charview so that the outlines remain visible behind them.
  • Make the background color of the debugger raster window configurable by the user
  • If a curved point had no control point in one direction, then ff thought it had no direction and felt free to change it. It should inherit the direction from the other side of the point. This was even more confusing if the point went through a two step process, first simplifying one side into a line (losing the cp) then simplifying the line away (losing the direction).
  • Use a better algorithm to indicate changed pixels. (in debug window)
  • Screwed up View->Show Grid Fit... in Mono mode.
  • Werner wants to be able to see grey scale rasters while debugging truetype. He also points out that if the instruction before the end of instructions changes pixels, ff would leave them marked "changed" even after finishing the instructions.
  • We were getting multiple error windows.
  • Don't allow users to add instructions to a glyph containing both a reference and contours. Or a glyph where a reference is scaled more than 200%. (tt doesn't allow these combinations so the references need to be copied inline)
  • Typo in remove overlap caused bad bug in feta26. Error introduced 15 Sept.
  • Further work on what makes a spline linear.
  • The utf7 output routine in sfd did not convert from utf8 properly.
  • Panov has found yet another error in simplify. Be really exuberant about turning splines that trace out lines into lines.
  • Panov finds another bug:
    • If we change the unicode value of a glyph, then we also need to change
    • the unicode value of any references to that glyph.
  • Panov presents four more bugs:
    • If an sfd file contains an unencoded glyph, it will crash
    • Force Encoding->Original will crash if there's a bdf font missing some characters (ie. the piecemeal display font)
    • Goto Dlg tried to free an uninit value
    • Goto Dlg contained a string initialized by latin1 rather than utf8.
  • Hmm. the changedsincelasthinted flag doesn't apply to truetype. The equivalent thing (I guess) is not having any instructions on a glyph with splines.
  • Panov wants a way to build up arrays using easier syntax than: create array, assign each element.
  • Sometimes we would get both an EUC-CN and a GB2312packed entry in the Encoding menu.
  • Create a plugin for GB12345, and treat it (in sfnt tables) as GB2312.
  • Add a new scripting funtion: NearlyLines to convert almost linear splines to linear.
  • Tweak AddExtrema so that it doesn't create Extrema points which are too close to the endpoints.
  • Didn't terminate an array of answer strings to a question dialog.
  • Fixed a crash in stem database.
• 28-Oct-2005
  • More work on the mac install procdure.
  • Redo the install documentation on the website & in the readmes
  • Add a hack to the directory browser so that on cygwin getting a directory listing of "/" will include the magic (fake) directory /cygdrive (which gives access to the rest of the PC)
  • Add a new scripting funtion: NearlyLines to convert almost linear splines to linear.
  • Tweak AddExtrema so that it doesn't create Extrema points which are too close to the endpoints.
  • I was dumping all opentype ligatures into an ofm(tfm) file. But I should probably only add 'liga' and 'rlig' ligatures.
  • Omega has a bug where it thinks 0 width glyphs do not exist. Which means it thinks none of the combining accents, etc. exist. So force the width of any zero width glyphs to be the smallest positive width.
  • When reading ttf/otf fonts, FF gave alternate substitutions (and multiple subs?) a tag of 0.
  • Add two scripting routines to allow people to read & write strings from and to files. One string per file.
  • The ClearTable scripting command parsed the table's tag incorrectly.
• 23-Oct-2005
  • Try to make the Mac install a bit more mac-like by adding an entry to the X11 Applications menu rather than expecting users to start fontforge from the command line.
  • David Binderman points out an out of bounds array reference in cvexport which has been there for years. Gleep.
  • Add scripting commands for manipulating ttf instructions.

    I added the following new scripting functions:

    •         SetMaxpValue("item-name", value)
    •         GetMaxpValue("item-name")
    •         ClearInstrs()
    •         ClearTable(tag)
    •         AddInstrs(thingamy,replace,instrs)
    •         FindOrAddCvtIndex(value[,sign-matters])
    •         ReplaceCvtAt(index,value)
    •         GetCvtAt(index)
    •         PrivateToCvt()
  • Daniel Gillmor points out there is no way to set the OS/2 Width field from a script. He then provided a patch with two new functions which does this. I realize there is no way to set any of the OS/2 values (except panose), and there should be (and to retrieve them too).

    I added the following new scripting functions:

    •         SetOS2Value("item-name", value)
    •         GetOS2Value("item-name")
    •         SetPrivateValue("item-key","item-value")
    •         GetPrivateValue("item-key")
    •         HasPrivateValue("item-key")
    •         RemovePrivateValue("item-key")
  • A number of fixes for ofm files.
  • When outputting mark-to-base lookups with multiple anchor classes, ff would put a base glyph in the coverage table several times if it was used by several mark classes. Bad.
• 18-Oct-2005
  • KANOU points out that I was using too strict a rule for parsing svg polygons.
  • OFM output was not working in fonts with lots of kerning combinations.
• 16-Oct-2005
  • A number of fixes for ofm files.
  • Add default 'nukt' "ligature" substitutions to appropriate Devanagari characters
• 15-Oct-2005
  • The patch to the routine determining what glyphs got output broke bitmap only fonts.
  • Ofm files didn't contain ligatures.
  • The ofm patch broke tfm output.
• 12-Oct-2005
  • Add support for generating & parsing ofm files (I hope)
  • If I tried to generate a font from an absolutely empty fontdb (ie. a new one) then I'd get a stupid error about how .notdef was at encoding -1 and this was a bad idea.
  • Add 4 new information types to scripting GlyphInfo

    * LayerCount

    Number of layers in glyph (usually 2)

    * RefCount

    Number of refs in glyph

    * RefNames

    Returns an array of all reference names (may be 0 length)

    * RefTransform

    Returns an array of all reference transforms

  • Don't mark glyphs with widths other than 1em as being worth outputting (require widthset). Used to use both (because in the early days, before pfaedit moved to sourceforge even, the widthset bit didn't exist). Now there are too many cases w here glyphs are created with a width other than 1em (marks, glyphs in monospace fonts (where the space is anything but 1em)).
  • When printing a type3 pdf font, give it a name. Not required, but so what.
  • Patch from Michal Nowakowski. When editing truetype instructions using my prefix notation for numbers, there was a crash if a number needed a word (rather than a byte).
  • Add some support for reading fonts out of a pdf file.
    • Limitted support for pdf type3 (no images)
    • Only support (Hex, 85, Flate, RLE)Decode filters (no support for lzw, fax, JBIG, DCT)
  • The routine to build the stem database (autohint) would set up a point database and then call a routine to convert all splines into a set of monotonics. But that routine could remove 0 length splines (and the points attached), which would mean our point db would be wrong.
  • Stefan Wanger suggests some patches to fix memory leaks.
  • Stefan Wanger gave me a type2 font which used type1 flex conventions (calling othersubrs, etc.) There is no othersubrs operator in type2, nor any place to store othersubrs subroutines. But it is easy enough to support, so why not. Warn that it's an illegal font, of course.
  • Stefan Wanger gave me a font containing the utf8 byte order mark in a fontname ( in a CFF Name Index). PostScript says FontNames are printable ASCII. But it's easy enough to skip over it.
  • Stepan Roh points out that in a monospace font, combining glyphs should default to the mono-width while in a proportional font they should default to 0 width (or perhaps 1 em-unit for compatability with old idiotic windows systems). Use the panose field to determine if a font is monospace.

    Actually ALL glyphs in a monospace font should default to the mono-space width rather than 1em.

  • New dialog: Hints->Edit 'maxp'... Gives user access to number of functions defined, the stack depth, storage usage, number of twilight points, which ff can't figure out for you.
  • Figured out how to make transparent images work on cygwin
  • Michal Nowakowski points out that negative values are duplicated in cvt.
  • Patch from Michal Nowakowski fixing crash when setting size of cvt table to 0.
• 29-Sept-2005
  • Add utf8 entry points for many (ucs2) routines. Minimal ability to draw non-BMP glyphs now with bdf fonts with encodings: *-UnicodePlane-1 (for plane 1), *-UnicodePlane-2 (for plane 2), etc.
  • Add a menu command to add values from the private dict into the cvt table (BlueValues, StemSnaps).
  • The Edit cvt window usually failed to draw anything in itself when it popped up (subsequent refreshes were ok).
  • Patch from Michal Nowakowski to improve memory usage in auto instr.
  • Printing sample text was broken in fonts with no kerning classes.
  • cygwin does very weird things with shared libaries, and we didn't support them at all.
  • FontForge did not support the depreciated usage of endchar to mean (almost) seac in type2 charstrings. It should now.
  • os2 typo linegap was not read out of sfd files.
  • Display 'gai' files in the open dlg.
  • I got some mnemonics wrong. Fix 'em up.
  • bdftopcf will use the glyph encoded at 0 as the default glyph if there is no explicit DEFAULT_CHAR. It seems happy with a DEFAULT_CHAR of -1 to mean no default, so use that if we don't want glyph 0 to be the default.
  • when outputting bdf, fnt or pt3 bitmap fonts with no outline font generated, we'd get filenames like foo-*-13.bdf. Get rid of the "-*".
  • Support point numbering of composites even when not debugging.
  • Michal Nowakowski points out that the cvt editor crashed if we changed the length of the table.
  • Provide two new scripting functions:
    • MoveReference(xoff,yoff,[ref-name/ref-unicode]+)
    • PositionReference(x,y,[ref-name/ref-unicode]+)

    These search all selected glyphs, looking for any references in those glyphs with the given name/unicode value, and then translating the reference by (xoff,yoff) or positioning the reference absolutely at (x,y).

  • ff failed to read horizontal metrics properly from a type2 CFF CID font. In most cases this would be masked by the fact that ff would correct the bad width values by reading the hmtx table. But if one had a bare CFF font, or a 'gai' font where there is no external source of widths, ff would get things wrong.
  • Alexej Kryukov points out that most entries in the private dict should be scaled when we do a ScaleToEm.
  • Fix a couple of problems with cid font display that "gai" fonts expose.
  • Michal Nowakowski points out that I generated truetype instructions --- and then forgot to attach them to the glyph. Whoops.
  • Loading extensible glyph information from a tfm file was broken by the encoding change.
• 19-Sept-2005
  • ff did not read AAT classes (kerning classes, etc) properly. It read one extra element which could cause problems
  • Cleanup svg output for kerning classes with no members.
  • Out of bounds array reference in OS2FigureCodePages caused a crash on the mac.
  • When loading a type1 font we failed to set the unicode encoding on seac references. This probably broke lots of things, it certainly broke replace with reference.
  • Add a DefaultRoundToGrid scripting function to set the Round-to-Grid truetype reference bit (basically set it whenever we have a reference which isn't point matched).
  • Add support for truetype point matching in references and anchor points
  • Improve the debugging heuristic that notices when we've jumpped to a new routine to work better with composites
  • If I attempted to debug a glyph with no instructions there was a race condition and