kmx git

Commit	Date	Message
47089fe7	2025-05-28T17:10:22	Big misc changes - Make geometry types into templates taking float or double. - Changes to paint() transform elimination. - Add font->has_nonzero_coords and use it across. - Extensions to avar processing for otherwise invalid mappings.
a9c2c635	2025-06-04T11:58:20	[sanitize] Fix pointer math Should fix https://github.com/harfbuzz/harfbuzz/issues/5347
633290dc	2025-03-03T14:51:09	[hvgl] Reuse sanitizer object
6ce9a207	2025-03-02T16:57:50	[hvgl] Hook up to hb-ot-font for get_path() Does nothing yet.
fc42cdd6	2025-01-08T22:17:03	[docs] Reduce warnings
d053426b	2023-11-06T11:44:14	Format
30672c7e	2023-11-03T18:52:03	Add hb_barrier() And annotate some more.
cf49f3f2	2023-11-03T18:20:14	[sanitize] Add barrier()
b16f1800	2023-11-03T17:51:50	[sanitize] Yet another hack to "fix" barrier issue
f380a328	2023-08-25T14:01:01	Fix some typos.
25e9defa	2023-07-31T17:12:07	[sanitize] Inline check_struct Though seems like the compiler was always inlining it anyway.
36c7ec44	2023-07-20T18:56:46	[sanitize] Simplify a return
837885f0	2023-07-20T12:58:16	Revert "[sanitize/Coverage] Keep a map of sane coverages" This reverts commit a689114898cc3e8f1c6ba7cc49cd6c3639d91250.
0ab90671	2023-07-20T12:57:16	[sanitize/Coverage] Keep a map of sane coverages Fonts like Gulzar reuse the same coverage over a thousand times sometimes. However, this doesn't speed up sanitize unfortunately. Looks like calling Coverage::sanitize() is already very fast. We're just doing A LOT of it. The map slowed it down in fact. A set was even slower. Going to revert.
253ec08b	2023-07-08T12:48:05	[sanitize] Speed up check_struct on x64
fb02f4a8	2023-07-08T12:29:18	[sanitize] Micro-optimize ChainRule::sanitize
73ee9c34	2023-07-07T22:47:20	Revert "[sanitize] Only check trailing edge of ranges" This reverts commit e4856cf8098b9741a910e7f7979096cf11d8fead. This is wrong on at least 32-bit systems. Might add a condition version later.
d8450420	2023-07-07T21:56:17	[sanitize] Optimize away an overflow check when not needed When the length argument is 16bit...
e4856cf8	2023-07-07T21:21:33	[sanitize] Only check trailing edge of ranges Assumptions... Speeds up Gulzar load_face_and_shape benchmark by 7%.
6a683ea6	2023-07-07T21:02:15	[sanitize] Enlighten check_range() some more No need to check for len=0 arrays. They must still be in range.
c24ea103	2023-07-07T20:43:20	[sanitize] Minor tweak to len=0 case which is handled otherwise anyway
59abcda2	2023-07-07T14:27:18	[sanitize] Add an ALWAYS_INLINE Looks like it was always inlined anyway.
5ac9eaa0	2023-07-02T16:18:08	[sanitize] Minor micro-optimize
bb33675e	2023-07-02T14:11:40	[sanitize] Micro-optimize check_range()
d1c82a1c	2023-07-02T13:38:22	[sanitize] Speed up check_struct() Don't account for length here. Should not be a problem. Fingers crossed fuzzers. 18% speed up in load_face_and_shape Gulzar-Regular.
2d6091fc	2023-06-30T11:48:56	[GPOS] Make AnchorMatrix sanitize lazy again Was reverted in the previous commit, because it was incomplete.
83eb744e	2023-06-30T11:33:39	[sanizie] Inline check_range if not OPTIMIZE_SIZE BM_Font/load_face_and_shape/NotoNastaliqUrdu-Regular.ttf/hb -0.1046 -0.1051 194 173 193 172 BM_Font/load_face_and_shape/NotoSerifMyanmar-Regular.otf/hb -0.2401 -0.2412 36 27 36 27
690af7aa	2023-06-30T10:36:01	[GPOS] Sanitize Device tables lazily This speeds up face loading for variable fonts by 80%! Comparing before to after Benchmark Time CPU Time Old Time New CPU Old CPU New --------------------------------------------------------------------------------------------------------------------------------------------------------------- BM_Font/load_face_and_shape/Roboto-Regular.ttf/hb -0.0368 -0.0366 20 20 20 19 BM_Font/load_face_and_shape/RobotoFlex-Variable.ttf/hb -0.7149 -0.7162 77 22 77 22 BM_Font/load_face_and_shape/RobotoFlex-Variable.ttf/var/hb -0.7241 -0.7255 80 22 79 22 BM_Font/load_face_and_shape/SourceSansPro-Regular.otf/hb -0.1441 -0.1445 28 24 28 24 BM_Font/load_face_and_shape/AdobeVFPrototype.otf/hb -0.7893 -0.7910 66 14 66 14 BM_Font/load_face_and_shape/AdobeVFPrototype.otf/var/hb -0.7865 -0.7882 67 14 66 14 BM_Font/load_face_and_shape/SourceSerifVariable-Roman.ttf/hb -0.8895 -0.8900 227 25 226 25 BM_Font/load_face_and_shape/SourceSerifVariable-Roman.ttf/var/hb -0.8895 -0.8900 226 25 225 25 BM_Font/load_face_and_shape/Comfortaa-Regular-new.ttf/hb -0.5512 -0.5531 42 19 42 19 BM_Font/load_face_and_shape/NotoNastaliqUrdu-Regular.ttf/hb -0.1511 -0.1510 227 192 225 191 BM_Font/load_face_and_shape/NotoSerifMyanmar-Regular.otf/hb -0.1494 -0.1498 41 35 40 34 OVERALL_GEOMEAN -0.6443 -0.6456 0 0 0 0
fe83736e	2023-03-02T11:35:42	[sanitize] Protect against an underflow
918193eb	2023-02-22T23:11:29	[subset] fix a class of fuzzer timeouts caused by large shared coverage tables. More acurately estimates the op count for CoverageFormat2 tables as the population size instead of the size in bytes.
6b72a4dd	2023-01-23T18:36:04	Revert "[sanitize] Simplify(?) check_range" This reverts commit af0b1ef8a72d4f6b778dbba3606ebe7df39d5288.
af0b1ef8	2023-01-23T14:31:10	[sanitize] Simplify(?) check_range
ed023f66	2023-01-12T17:04:24	Enable -Wformat-signedness And fix the codebase.
4c14043b	2022-11-21T12:56:33	[algs] Add output argument to hb_unsigned_mul_overflows()
64b29dbd	2021-11-09T09:13:14	[subset] reuse colrv1 max nesting depth constant for sanitize.
ace98cc6	2021-11-08T15:47:56	[subset] Only sanitize recursion depth in COLR.
782a7377	2021-11-04T13:39:11	[sanitize] add a maximum recursion depth for sanitize.
6d555ce8	2021-11-02T00:18:22	[meta] Use std::forward instead of hb_forward()
8ef4257d	2021-03-31T15:00:46	[sanitize] change max ops to track number of bytes processed. Counting bytes as the operations is likely to be a better proxy for how much work processing the table will cost vs. the current approach of counting the number of sub-objects. This should allow checks for max features, max scripts, etc. to be removed. I tested this change against the full collection of fonts at https://github.com/google/fonts and a max ops factor of 3 was sufficient to successfully sanitize all of them.
a650243d	2021-01-24T13:30:06	[sanitize] Fix typo
a8e72ee7	2020-12-30T23:08:40	[docs] Use %true and %false consistently
1535440b	2020-06-19T08:30:59	[dispatch] Fix debug builds
b398748d	2020-03-04T11:18:19	[algs] Add hb_clamp Similar to stl and glsl's clamp
d3836039	2020-03-02T22:41:08	Limit OT::Lookup subtables (#2219) Fixes https://crbug.com/oss-fuzz/13943
b28c2825	2019-12-17T02:58:51	Check to avoid overflows
d70afb1e	2019-12-17T02:29:28	Clamp max_ops to upper bound in hb-sanitize.hh
a0b4ac4d	2019-08-24T17:57:14	Turn 8 spaces to tab across the project According to the current code style of the project
25a5b287	2019-05-10T16:01:39	Fix sanitize fail of extension sublookups Fixes https://bugs.chromium.org/p/chromium/issues/detail?id=960331
2ba984fc	2019-05-07T23:28:22	Fix signed comparison on 32bit
83e3eabd	2019-05-07T20:58:43	Whitespace
41248cce	2019-05-07T20:54:31	Remove MIN/MAX in favor of hb_min/hb_max
c14efb8e	2019-05-05T09:54:58	Fix previous commit Priority should be given to specific over dispatch. Broke sanitize before. This fixes it, by moving prioritization to the context implementation, since the correct priority cannot be done in the dispatch implementation. Done for subset and sanitize only, which need it.
b10f6593	2019-05-05T09:23:35	[dispatch] Use functionality from previous commit To remove a couple of unwanted wrapper methods
ac350c92	2019-05-05T09:10:46	[dispatch] Try obj.dispatch(c) before trying c->dispatch(obj)
a7c63cd8	2019-03-30T14:59:40	Split sanitize and dispatch into their own files

47089fe7

2025-05-28T17:10:22

Big misc changes - Make geometry types into templates taking float or double. - Changes to paint() transform elimination. - Add font->has_nonzero_coords and use it across. - Extensions to avar processing for otherwise invalid mappings.

a9c2c635

2025-06-04T11:58:20

[sanitize] Fix pointer math Should fix https://github.com/harfbuzz/harfbuzz/issues/5347

633290dc

2025-03-03T14:51:09

[hvgl] Reuse sanitizer object

6ce9a207

2025-03-02T16:57:50

[hvgl] Hook up to hb-ot-font for get_path() Does nothing yet.

fc42cdd6

2025-01-08T22:17:03

[docs] Reduce warnings

d053426b

2023-11-06T11:44:14

Format

30672c7e

2023-11-03T18:52:03

Add hb_barrier() And annotate some more.

cf49f3f2

2023-11-03T18:20:14

[sanitize] Add barrier()

b16f1800

2023-11-03T17:51:50

[sanitize] Yet another hack to "fix" barrier issue

f380a328

2023-08-25T14:01:01

Fix some typos.

25e9defa

2023-07-31T17:12:07

[sanitize] Inline check_struct Though seems like the compiler was always inlining it anyway.

36c7ec44

2023-07-20T18:56:46

[sanitize] Simplify a return

837885f0

2023-07-20T12:58:16

Revert "[sanitize/Coverage] Keep a map of sane coverages" This reverts commit a689114898cc3e8f1c6ba7cc49cd6c3639d91250.

0ab90671

2023-07-20T12:57:16

[sanitize/Coverage] Keep a map of sane coverages Fonts like Gulzar reuse the same coverage over a thousand times sometimes. However, this doesn't speed up sanitize unfortunately. Looks like calling Coverage::sanitize() is already very fast. We're just doing A LOT of it. The map slowed it down in fact. A set was even slower. Going to revert.

253ec08b

2023-07-08T12:48:05

[sanitize] Speed up check_struct on x64

fb02f4a8

2023-07-08T12:29:18

[sanitize] Micro-optimize ChainRule::sanitize

73ee9c34

2023-07-07T22:47:20

Revert "[sanitize] Only check trailing edge of ranges" This reverts commit e4856cf8098b9741a910e7f7979096cf11d8fead. This is wrong on at least 32-bit systems. Might add a condition version later.

d8450420

2023-07-07T21:56:17

[sanitize] Optimize away an overflow check when not needed When the length argument is 16bit...

e4856cf8

2023-07-07T21:21:33

[sanitize] Only check trailing edge of ranges Assumptions... Speeds up Gulzar load_face_and_shape benchmark by 7%.

6a683ea6

2023-07-07T21:02:15

[sanitize] Enlighten check_range() some more No need to check for len=0 arrays. They must still be in range.

c24ea103

2023-07-07T20:43:20

[sanitize] Minor tweak to len=0 case which is handled otherwise anyway

59abcda2

2023-07-07T14:27:18

[sanitize] Add an ALWAYS_INLINE Looks like it was always inlined anyway.

5ac9eaa0

2023-07-02T16:18:08

[sanitize] Minor micro-optimize

bb33675e

2023-07-02T14:11:40

[sanitize] Micro-optimize check_range()

d1c82a1c

2023-07-02T13:38:22

[sanitize] Speed up check_struct() Don't account for length here. Should not be a problem. Fingers crossed fuzzers. 18% speed up in load_face_and_shape Gulzar-Regular.

2d6091fc

2023-06-30T11:48:56

[GPOS] Make AnchorMatrix sanitize lazy again Was reverted in the previous commit, because it was incomplete.

83eb744e

2023-06-30T11:33:39

[sanizie] Inline check_range if not OPTIMIZE_SIZE BM_Font/load_face_and_shape/NotoNastaliqUrdu-Regular.ttf/hb -0.1046 -0.1051 194 173 193 172 BM_Font/load_face_and_shape/NotoSerifMyanmar-Regular.otf/hb -0.2401 -0.2412 36 27 36 27

690af7aa

2023-06-30T10:36:01

[GPOS] Sanitize Device tables lazily This speeds up face loading for variable fonts by 80%! Comparing before to after Benchmark Time CPU Time Old Time New CPU Old CPU New --------------------------------------------------------------------------------------------------------------------------------------------------------------- BM_Font/load_face_and_shape/Roboto-Regular.ttf/hb -0.0368 -0.0366 20 20 20 19 BM_Font/load_face_and_shape/RobotoFlex-Variable.ttf/hb -0.7149 -0.7162 77 22 77 22 BM_Font/load_face_and_shape/RobotoFlex-Variable.ttf/var/hb -0.7241 -0.7255 80 22 79 22 BM_Font/load_face_and_shape/SourceSansPro-Regular.otf/hb -0.1441 -0.1445 28 24 28 24 BM_Font/load_face_and_shape/AdobeVFPrototype.otf/hb -0.7893 -0.7910 66 14 66 14 BM_Font/load_face_and_shape/AdobeVFPrototype.otf/var/hb -0.7865 -0.7882 67 14 66 14 BM_Font/load_face_and_shape/SourceSerifVariable-Roman.ttf/hb -0.8895 -0.8900 227 25 226 25 BM_Font/load_face_and_shape/SourceSerifVariable-Roman.ttf/var/hb -0.8895 -0.8900 226 25 225 25 BM_Font/load_face_and_shape/Comfortaa-Regular-new.ttf/hb -0.5512 -0.5531 42 19 42 19 BM_Font/load_face_and_shape/NotoNastaliqUrdu-Regular.ttf/hb -0.1511 -0.1510 227 192 225 191 BM_Font/load_face_and_shape/NotoSerifMyanmar-Regular.otf/hb -0.1494 -0.1498 41 35 40 34 OVERALL_GEOMEAN -0.6443 -0.6456 0 0 0 0

fe83736e

2023-03-02T11:35:42

[sanitize] Protect against an underflow

918193eb

2023-02-22T23:11:29

[subset] fix a class of fuzzer timeouts caused by large shared coverage tables. More acurately estimates the op count for CoverageFormat2 tables as the population size instead of the size in bytes.

6b72a4dd

2023-01-23T18:36:04

Revert "[sanitize] Simplify(?) check_range" This reverts commit af0b1ef8a72d4f6b778dbba3606ebe7df39d5288.

af0b1ef8

2023-01-23T14:31:10

[sanitize] Simplify(?) check_range

ed023f66

2023-01-12T17:04:24

Enable -Wformat-signedness And fix the codebase.

4c14043b

2022-11-21T12:56:33

[algs] Add output argument to hb_unsigned_mul_overflows()

64b29dbd

2021-11-09T09:13:14

[subset] reuse colrv1 max nesting depth constant for sanitize.

ace98cc6

2021-11-08T15:47:56

[subset] Only sanitize recursion depth in COLR.

782a7377

2021-11-04T13:39:11

[sanitize] add a maximum recursion depth for sanitize.

6d555ce8

2021-11-02T00:18:22

[meta] Use std::forward instead of hb_forward()

8ef4257d

2021-03-31T15:00:46

[sanitize] change max ops to track number of bytes processed. Counting bytes as the operations is likely to be a better proxy for how much work processing the table will cost vs. the current approach of counting the number of sub-objects. This should allow checks for max features, max scripts, etc. to be removed. I tested this change against the full collection of fonts at https://github.com/google/fonts and a max ops factor of 3 was sufficient to successfully sanitize all of them.

a650243d

2021-01-24T13:30:06

[sanitize] Fix typo

a8e72ee7

2020-12-30T23:08:40

[docs] Use %true and %false consistently

1535440b

2020-06-19T08:30:59

[dispatch] Fix debug builds

b398748d

2020-03-04T11:18:19

[algs] Add hb_clamp Similar to stl and glsl's clamp

d3836039

2020-03-02T22:41:08

Limit OT::Lookup subtables (#2219) Fixes https://crbug.com/oss-fuzz/13943

b28c2825

2019-12-17T02:58:51

Check to avoid overflows

d70afb1e

2019-12-17T02:29:28

Clamp max_ops to upper bound in hb-sanitize.hh

a0b4ac4d

2019-08-24T17:57:14

Turn 8 spaces to tab across the project According to the current code style of the project

25a5b287

2019-05-10T16:01:39

Fix sanitize fail of extension sublookups Fixes https://bugs.chromium.org/p/chromium/issues/detail?id=960331

2ba984fc

2019-05-07T23:28:22

Fix signed comparison on 32bit

83e3eabd

2019-05-07T20:58:43

Whitespace

41248cce

2019-05-07T20:54:31

Remove MIN/MAX in favor of hb_min/hb_max

c14efb8e

2019-05-05T09:54:58

Fix previous commit Priority should be given to specific over dispatch. Broke sanitize before. This fixes it, by moving prioritization to the context implementation, since the correct priority cannot be done in the dispatch implementation. Done for subset and sanitize only, which need it.

b10f6593

2019-05-05T09:23:35

[dispatch] Use functionality from previous commit To remove a couple of unwanted wrapper methods

ac350c92

2019-05-05T09:10:46

[dispatch] Try obj.dispatch(c) before trying c->dispatch(obj)

a7c63cd8

2019-03-30T14:59:40

Split sanitize and dispatch into their own files

kc3-lang/harfbuzz/src/hb-sanitize.hh

src/hb-sanitize.hh

Log