6215fb8e68
[serialize] Actually reclaim storage from duplicate objects
2019-04-02 23:10:03 -07:00
3305a2cad2
[serialize] Port to use object pool
...
Tested, but feels fragile :(.
2019-04-02 22:58:49 -07:00
5efbc01174
[pool] Uses memset() instead of assigning Null()
...
Assignment is invalid on invalid object.
2019-04-02 22:41:06 -07:00
434d78bf91
Add hb_pool_t<> for pooled memory allocation
2019-04-02 21:46:40 -07:00
8e4df1a152
[serialize] Disable packed_map again
...
Ugh. Need to think of something else.
2019-04-02 20:20:53 -07:00
31c1a83899
[map] Protect more against pointer deref
2019-04-02 20:17:27 -07:00
5bffa9e375
More
2019-04-02 20:13:16 -07:00
5b66b033fd
[serialize] Fix hb_hashmap_t<> for pointers and use in packed_map
2019-04-02 19:30:22 -07:00
42ab32cbba
[iter] Remove passing pointer to hb_iter()
...
While doable with hb_deref_pointer() as well, we also would then
need to do it in a ton of places. Not worth it / messy.
2019-04-02 18:42:51 -07:00
d0da547b37
[array] Use dagger for hashing array
...
Also switch to better mixing.
2019-04-02 18:23:05 -07:00
b6f29bf141
[iter] Accept pointers in hb_iter()
...
No idea how to avoid dupicating code. Was hoping hb_deref_pointer()
would do it, but looks like a pointer can't bind to a universal
reference T&&. Humm.
2019-04-02 18:12:01 -07:00
fc24bb9046
[serialize] Towards maintaining packed_map
2019-04-02 17:49:52 -07:00
aa2293a55e
[serialize] Minor
2019-04-02 17:42:10 -07:00
e42b82c828
[serialize] Handle non-nullable offsets
2019-04-02 17:23:16 -07:00
e04518bafc
[serialize] Movce empty-object handling earlier
2019-04-02 17:20:04 -07:00
7f73c9744e
[serialize] Minor
2019-04-02 17:12:24 -07:00
5a3de4f4f8
[serialize] Allow offset links that have base offset from the object base
...
Rarely used, but used, in name table or similar constructs.
2019-04-02 16:53:42 -07:00
edad6b2c45
[test][iter] minor, fix double promotion warning
2019-04-03 00:48:59 +04:30
062f5d6e7a
[test] minor, c style comments
2019-04-02 20:27:00 +04:30
afdbf960d6
[iter][test] Add another test for hb_reduce
...
Different initial and accumulator types
2019-04-02 20:25:06 +04:30
bfa02bef45
[serialize] Switch to tetris-packing
2019-04-01 21:36:13 -07:00
f0ea3ac17b
[serialize] Fix linking
2019-04-01 21:36:03 -07:00
7c0e2054e0
[serialize] Minor
2019-04-01 21:32:29 -07:00
64d0f08938
[cmap] Minor
2019-04-01 16:56:41 -07:00
2e675cc7b5
[subset] Call serialize start/end around it
...
To be cleaned up.
2019-04-01 16:56:41 -07:00
abe33c6149
[serialize] Assert stack
2019-04-01 16:56:41 -07:00
72e9b2c16c
[serialize] Add add_link() to add link
2019-04-01 16:56:41 -07:00
74addbecac
[serialize] Add default template type to push()
2019-04-01 16:56:41 -07:00
b8642087e6
[iter] hb_reduce, accumulator with a different type
2019-04-01 16:56:29 -07:00
e526414c75
[iter] Implement hb_reduce
2019-04-01 16:56:29 -07:00
f3aca6aa26
[serialize] Implement linking
...
Untested!
2019-03-31 21:37:14 -07:00
17f0cfa7ea
Move BEInt to hb.hh
...
I knows...
2019-03-31 21:34:19 -07:00
78fc43f293
[iter] Fix up build, ouch
...
Yeah, some things not very clear...
2019-03-31 19:17:07 -07:00
e5d6fe9782
Don't skip setting the .end field of the first range
...
Fixes a bug in CoverageFormat2::serialize whereby the first range
was not serialized correctly if it consists of only a single glyph ID.
This broke shaping of U+0626 in the Arabic fallback shaper, because it
is not found in the coverage table of the 'init' and 'medi' lookups.
Also fix similar bug in ClassDefFormat2::serialize, noted during code
inspection (I haven't observed a case that was actually affected by
this, but it looks broken).
Fixes https://github.com/harfbuzz/harfbuzz/issues/1504
2019-03-31 19:02:47 -07:00
8a8d45b924
[iter] Adjust hb_copy() and use it
...
Untested.
2019-03-31 19:00:09 -07:00
ba4b7be455
Remove coretext_aat shaper ( #1581 )
...
coretext_aat was a temporary shaper to redirect shaping of AAT fonts
to CoreText and leaving the rest for HarfBuzz. As HarfBuzz now supports
AAT and Chrome now actually ships that on a stable version on macOS,
we no longer care about such use-case. If a client really wants 100%
metrics compatibility with CoreText better to use it directly or through
our API. Replicating the same behavior still is possible using
hb_shape_full, something we don't care or like to offer anymore.
Fixes https://github.com/harfbuzz/harfbuzz/issues/1478
2019-03-31 01:32:30 -07:00
d6005b49b3
[serialize] Start implementing linking
2019-03-30 19:49:56 -07:00
313b3057c3
[serializer] Implement dedup!
2019-03-30 19:46:35 -07:00
b189bbc48f
Implement hashing of objects
...
Should be improved for hb_bytes_t.
2019-03-30 19:41:48 -07:00
d6b28057a5
Fix hb_hash(pointer)
2019-03-30 19:31:51 -07:00
d74dc3ef65
[serialize] Don't insert empty object into tree
2019-03-30 19:26:37 -07:00
7fd8228326
[serialize] Towards maintaining hashmap
2019-03-30 19:16:20 -07:00
f254f45a1e
[serialize] Only pack main object if there are other objects
...
Avoids a memmove for tables that don't use the object packing mechanism.
2019-03-30 19:05:51 -07:00
946d446f9b
[serialize] Copy both sides of the buffer
2019-03-30 19:03:55 -07:00
8512dc565d
[serialize] Simplify copy
2019-03-30 19:01:23 -07:00
0b1fe7b716
[serializer] Unbreak for now
2019-03-30 18:48:26 -07:00
10f062234e
[map] Shuffle fini code
2019-03-30 18:44:01 -07:00
e6b78003ef
[vector] Add move semantics
2019-03-30 18:33:30 -07:00
4c4d3c3ed5
[vector] Add some move and forwarding
2019-03-30 18:30:50 -07:00
7c9ceabcef
[meta] Add hb_move and hb_forward ala std::
2019-03-30 18:23:24 -07:00
Behdad Esfahbod
Ebrahim Byagowi
Jonathan Kew