Load SkiaRegular.ttf (version 4.0) makes FontForge crash.

[lemzwerg]

[Based on sources from 22:50 CET 8-Mar-2015-ML-TtfDb-D. Based on source from git with hash: e27e285fb052cc52ca5548103494ace48563aecd]

Calling fontforge SkiaRegular.ttf causes the following assertion failure.

fontforge: malloc.c:2369: sysmalloc:
Assertion `(old_top == (((mbinptr) (((char *) &((av)->bins[((1) - 1) * 2])) -
__builtin_offsetof (struct malloc_chunk, fd)))) && old_size == 0) ||
((unsigned long) (old_size) >= (unsigned long)((((__builtin_offsetof (
struct malloc_chunk, fd_nextsize))+((2 *(sizeof(size_t))) - 1)) &
~((2 *(sizeof(size_t))) - 1))) && ((old_top)->size & 0x1) &&
((unsigned long) old_end & pagemask) == 0)' failed.

This is a regression. FontForge version 31-Jul-2012 loads the font.

[adrientetar]

I don't have access to this font. Can you post a backtrace?

[lemzwerg]

I can't, since I currently don't have a debug build. However, give me an e-mail address, and I will send you the font privately.

[lemzwerg]

OK, sent.

[lemzwerg]

Oops! I get

SMTP error from remote server after RCPT command:
host: mx2.hotmail.com
Requested action not taken: mailbox unavailable

Do you have another e-mail address?

[lemzwerg]

OK, resent.

[adrientetar]

There are some warnings but it does not crash for me.

[lemzwerg]

Hmm. I'm compiling with

./bootstrap
./configure --enable-debug-raw-points \
            --enable-silent-rules \
            --enable-freetype-debugger=/path/to/freetype \
            CFLAGS="-g -O0"

on a 32bit GNU/Linux box using gcc 4.8.3. Maybe a 32bit vs. 64bit issue...

[frank-trampe]

@lemzwerg, send me the file.

[lemzwerg]

Will do. What e-mail address shall I use?

[frank-trampe]

I just sent you an e-mail.

[lemzwerg]

Thanks. Font sent.

[frank-trampe]

@lemzwerg, I figured out what is happening, but I'm not sure how best to deal with it.

InfoCopyGlyphs makes the glyphs for the variations, but it only allocates according the original/actual glyph count. _readttfmort then increases the glyph count and adds space in the main glyph storage info->chars (but not in the variation glyph storage) for fake glyphs. This causes problems when dealing with the variations, because the main glyph count is the only measure of how many spaces are allocated there and may be larger than the actual number of spaces allocated, causing a crash in SFFromTuple.

There are a few possible options.

• We could have each variation store a count of its contained glyphs.
• We could have a new count in ttfinfo for real glyphs.
• We could use info->badgids[0]->orig_pos if it exists as the end of the real glyph space.
• We could go back to the variations and extend their storage space whenever changing info->glyph_cnt.

I lack sufficient knowledge of TrueType to determine which of these is best. Any thoughts?

[lemzwerg]

Unfortunately, my knowledge of GX fonts is still quite limited (and I have zero knowledge regarding GX font editing). @twardoch, can you make a suggestion here?