9db42c09e9
Roll new brotli version.
2016-12-06 11:14:53 +01:00
cdd850ef52
Don’t ignore write failure
...
Makes it easy to find the source of the problems than failing later.
2016-08-09 15:59:50 +02:00
8f3ff26c37
Fix an out of bounds access found by libFuzzer
2016-05-17 16:00:38 -07:00
a15a8ab86b
Fixes for bugs uncovered by libFuzzer analysis. A write past buffer, a potential DOS by way of giant alloc, and an out of range access.
2016-05-06 13:14:42 -07:00
2bc6acf6df
Push fix for https://bugs.chromium.org/p/chromium/issues/detail?id=604649
2016-04-21 09:29:50 -07:00
07cd303dd2
Update to latest brotli. Use size_t more consistently. Remove dead code.
2016-04-21 09:27:28 -07:00
b220918cab
Stop trusting header totalSfntSize and glyf origLength
2016-03-08 12:30:08 -08:00
d1efde9124
Update woff2_dec.cc range checking
2016-02-04 08:24:36 -08:00
14f4cb67ec
Tweaks requested by @fred-wang to help Mozilla build
2016-01-28 11:34:31 -08:00
c817c969bf
hmtx optimization decoding. never encodes hmtx optimized as yet; need a cli param.
2016-01-20 14:10:02 -08:00
965d077bde
Fix signed/unsigned comparison warnings on MSVC
2015-11-17 18:20:03 +09:00
17ed658d0a
Don’t print to stderr unless a standalone binary
...
We will use woff2 code from OTS and it is undesirable for a library to
print messages to stderr.
2015-08-28 01:30:03 +02:00
eeec219f20
Only reorder the table directory
2015-07-27 08:47:58 -07:00
208c8f854e
Kill kWoff2FlagsContinueStream
...
There is only a single data stream for all the compressed tables, but
the code was originally written for separately compressed tables then
retrofitted, so this is an attempt to streamline it a bit.
2015-07-15 11:25:52 +02:00
4e3bb3418f
Merge PR #16 (reject composite glyphs w/o bbox), slightly modified to match recent code updates (mostly just adding branch predictions)
2015-07-14 10:12:40 -07:00
ea4c87a2c5
Updates from internal copy. Better branch prediction, ability to set Brotli quality on ConvertTTFToWOFF2
2015-07-13 09:08:04 -07:00
c54533f47f
Fix compiler warning on Mac OS X
...
warning: format specifies type 'unsigned long' but the argument has type
'uint64_t' (aka 'unsigned long long') [-Wformat]
2015-06-25 03:32:08 +02:00
531adab1fe
Reject non-zero loca transformLength
2015-04-20 00:21:05 +02:00
3f91c6e58f
Check matadata and private blocks offsets & length
...
So that we can reject fonts with extraneous data between blocks, as the
spec says we should reject says we should reject such fonts.
2015-04-20 00:11:04 +02:00
0634e0a719
Don't accept too long totalSfntSize
...
Per the spec (WOFF 2.0 spec just refer to WOFF 1.0 spec here):
http://www.w3.org/TR/2012/REC-WOFF-20121213/#conform-totalsize-longword-reject
2015-04-20 00:11:04 +02:00
498c637024
Correct decoder output table ordering
2015-03-03 09:30:12 -08:00
1c975530b9
Adding TTC support, take 1
2015-02-24 13:53:35 -08:00
445f541996
Sort tables when decoding
2015-02-12 09:56:29 -08:00
2b35420fa9
As suggested in https://github.com/google/woff2/pull/2 , correct const for top 7 bits
2014-11-24 07:11:04 -08:00
09c4eba0f6
Rename woff2 to src
2014-11-22 16:35:21 +02:00
Eugene Kliuchnikov