1692220 - New long-ish delay between first paint and presenting initial about:home due to content-process sandbox blocking font access

Status: RESOLVED FIXED

(Toolkit :: Startup and Profile System, defect)

Description

[Jonathan Kew [:jfkthame]]

+++ This bug was initially created as a clone of Bug #1690877 +++

This is a continuation of bug 1690877.

As noted in bug 1690877 comment 56, the patch that landed there has improved the performance significantly; however, at least on macOS 10.15 we are still taking a hit apparently as a result of the content-process sandboxing.

We should probably poke an appropriate hole in the sandbox to ensure the content process can directly access the libFontRegistry caches to avoid this.

Comment 1

[Jonathan Kew [:jfkthame]]

I've pushed a try build with a sandboxing patch based on Haik's from bug 1690877 comment 36, but restricted to just the com.apple.FontRegistry directory which is my current guess as to the one we need.
https://treeherder.mozilla.org/#/jobs?repo=try&revision=13ad023619a2db6cefd5dc5008be851bacd92000

Comment 2

[Jonathan Kew [:jfkthame]]

:jesup, when you have a chance to test the build from comment 1, it'd be good to know if it is sufficient to avoid the problem. Thanks!

Comment 3

[Randell Jesup [:jesup] (needinfo me)]

Per matrix, it did avoid my problem with 2s+ process startup

Comment 4

[Jonathan Kew [:jfkthame]]

Attachment: Bug 1692220 - Allow content-process read access to libFontRegistry caches. r=haik

Comment 5

[Haik Aftandilian [:haik]]

Attachment: Bug 1692220 - Add a test to ensure the com.apple.FontRegistry dir is readable from content processes r?jfkthame!

Comment 6

[Haik Aftandilian [:haik]]

Clean try push for the test: https://treeherder.mozilla.org/jobs?repo=try&selectedTaskRun=T_u7BOU9Q46zeXIi8WRfZA.0&revision=6e7bc409472c6f094d92ac26ab37a1645ef07f70

The test asserts that the registry dir and a file within it named font have already been created by the system. This appears to be true on our automation machines. If this turns out to be a bad assumption, the test will fail and we can change the test to not require this.

Comment 7

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/02155ea4c6f7
Allow content-process read access to libFontRegistry caches. r=haik
https://hg.mozilla.org/integration/autoland/rev/58f29cf068ae
Add a test to ensure the com.apple.FontRegistry dir is readable from content processes r=jfkthame

Comment 8

[Cristian Brindusan [:cbrindusan]]

https://hg.mozilla.org/mozilla-central/rev/02155ea4c6f7
https://hg.mozilla.org/mozilla-central/rev/58f29cf068ae