1476334 - Update HarfBuzz to 1.8.7

Status: RESOLVED FIXED

(Core :: Graphics: Text, defect, P3)

Description

[Ryan VanderMeulen [:RyanVM]]

Overview of changes leading to 1.8.4
Tuesday, July 17, 2018
====================================
- Fix build on non-C++11.
- Use C++-style GCC atomics and C++11 atomics.

Not sure if there's much urgency to take this update, but at least we've got a bug filed for it now.

Comment 1

[NVD]

https://cgit.freedesktop.org/harfbuzz/commit/?id=44d1fb37efa20852cc466c0f0bba95dbd24ce288

Overview of changes leading to 1.8.5
Wednesday, August 1, 2018
====================================
- Major Khmer shaper improvements to better match Microsoft.
- Indic bug fixes.
- Internal improvements to atomic operations.

Comment 2

[Ryan VanderMeulen [:RyanVM]]

Overview of changes leading to 1.8.6
Tuesday, August 7, 2018
====================================
- Internal code shuffling.
- New API to speed up getting advance widths for implementations
  that have heavy overhead in get_h_advance callback:
+hb_font_funcs_set_glyph_h_advances_func
+hb_font_funcs_set_glyph_v_advances_func
+hb_font_get_glyph_advances_for_direction
+hb_font_get_glyph_h_advances
+hb_font_get_glyph_h_advances_func_t
+hb_font_get_glyph_v_advances
+hb_font_get_glyph_v_advances_func_t

Comment 3

[Ryan VanderMeulen [:RyanVM]]

I had a green Try run with 1.8.5 yesterday, but 1.8.6 appears to have a new reftest hang:
https://treeherder.mozilla.org/logviewer.html#?job_id=192590549&repo=try

Comment 4

[Ryan VanderMeulen [:RyanVM]]

Also assertions:
https://treeherder.mozilla.org/logviewer.html#?job_id=192596671&repo=try

Comment 5

[David Baron :dbaron: (⌚️UTC-4, no longer working on Mozilla)]

The 1.8.5 update sounds valuable from a bug fix / language support perspective.

Comment 6

[Jonathan Kew [:jfkthame]]

Ugh... looks like the reftest failure also starts with a similar assertion, actually:
https://treeherder.mozilla.org/logviewer.html#?job_id=192590549&repo=try&lineNumber=29902

Behdad, does this look like an internal harfbuzz issue, or like we're just using it wrong? (Although it didn't complain until now!)

Comment 7

[Ryan VanderMeulen [:RyanVM]]

Yeah, it is indeed debug-only. The opt runs are all green.

Comment 8

[Behdad Esfahbod]

Oops.  My bug.  Fixed:
https://github.com/harfbuzz/harfbuzz/commit/883957688cc4aa4335e5131a60b3ac3b53290726

I'll get a release out.

Comment 9

[Behdad Esfahbod]

1.8.7 is out.

Comment 10

[Francois Guerraz]

1.8.6 rolled out on arch linux and makes Nightly crash when loading many pages, like https://xxx.onelogin.com/login

Crashreport: https://crash-stats.mozilla.com/report/index/3be1fb4a-57a8-41f0-ab56-3cdb60180809

Comment 11

[Francois Guerraz]

looks like I should upgrade to 1.8.7 :D

Comment 12

[Jonathan Kew [:jfkthame]]

I pushed a try run with 1.8.7 to confirm that it no longer asserts after the upstream fix:
https://treeherder.mozilla.org/#/jobs?repo=try&revision=f1a27fbf1ab3e7048a1fc10d49cc8c466c162fc4

Comment 13

[Ryan VanderMeulen [:RyanVM]]

Overview of changes leading to 1.8.7
Wednesday, August 8, 2018
====================================
- Fix assertion failure with GDEF-blacklisted fonts.

Comment 14

[Ryan VanderMeulen [:RyanVM]]

Attachment: Update HarfBuzz to version 1.8.7

Your Try run looks good.

Comment 15

[Jonathan Kew [:jfkthame]]

Comment on attachment 8998829 [details] [diff] [review]
Update HarfBuzz to version 1.8.7

Yeah, looks better -- let's go for it.

Comment 16

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/b0cdb1cc6d65
Update HarfBuzz to version 1.8.7. r=jfkthame

Comment 17

[Noemi Erli[:noemi_erli]]

https://hg.mozilla.org/mozilla-central/rev/b0cdb1cc6d65

Comment 18

[Behdad Esfahbod]

Does Firefox use hb_icu_get_unicode_funcs()?  If yes, please update to 1.8.8, as that contained a crasher on ARM introduced in 1.8.4.

Comment 19

[Ryan VanderMeulen [:RyanVM]]

The only appearance in our codebase appears to be in HarfBuzz itself, so I think we're safe.

Comment 20

[Jonathan Kew [:jfkthame]]

Right; we provide our own unicode_funcs, and don't use (or even build) the hb_icu support.