Bug 1516540 Comment 15 Edit History

Note: The actual edited comment in the bug view page will always show the original commenter’s name and original timestamp.

So I have some more interesting updates to share here.

Firstly, this is where we are currently standing, with all of the fixes that have landed so far: <https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=1b73a3b810a4b8edab6451c106a866429fc59d95&newProject=try&newRevision=15313fe8f75a63db2264ea8391c8f9342c6bb7a5&framework=10&showOnlyImportant=1>  Things are still not looking good.  Although I suspect bug 1270059 has sort of changed the baselines here quite significantly.  I'm not actually sure how meaningful it is to compare the geometric means of the subtests before/after that change any more.

I decided to follow an approach of eliminating possible culprits, by pushing trial patches to the try server that progressively disabled code that is hidden behind the checks for network.cookie.cookieBehavior being equal to 4, especially now that I discovered bug 1523705.

Look and behold.  This patch <https://hg.mozilla.org/try/rev/80f83ae8e6284dd7c71ac6772c92a06697f43e1f> erases all of the regressions: <https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=1b73a3b810a4b8edab6451c106a866429fc59d95&newProject=try&newRevision=a08c3e5f54a7c14017a1c259181496ecb34666c0&framework=10&showOnlyImportant=1>.  IOW, all of the regressions are coming loading images...
So I have some more interesting updates to share here.

Firstly, this is where we are currently standing, with all of the fixes that have landed so far: <https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=1b73a3b810a4b8edab6451c106a866429fc59d95&newProject=try&newRevision=15313fe8f75a63db2264ea8391c8f9342c6bb7a5&framework=10&showOnlyImportant=1>  Things are still not looking good.  Although I suspect bug 1270059 has sort of changed the baselines here quite significantly.  I'm not actually sure how meaningful it is to compare the geometric means of the subtests before/after that change any more.

I decided to follow an approach of eliminating possible culprits, by pushing trial patches to the try server that progressively disabled code that is hidden behind the checks for network.cookie.cookieBehavior being equal to 4, especially now that I discovered bug 1523705.

Look and behold.  This patch <https://hg.mozilla.org/try/rev/80f83ae8e6284dd7c71ac6772c92a06697f43e1f> erases all of the regressions: <https://treeherder.mozilla.org/perf.html#/compare?originalProject=try&originalRevision=1b73a3b810a4b8edab6451c106a866429fc59d95&newProject=try&newRevision=a08c3e5f54a7c14017a1c259181496ecb34666c0&framework=10&showOnlyImportant=1>.  IOW, all of the regressions are coming from loading images in this codepath: <https://searchfox.org/mozilla-central/rev/78cd247b5d7a08832f87d786541d3e2204842e8e/toolkit/components/antitracking/AntiTrackingCommon.cpp#1341-1401>

Back to Bug 1516540 Comment 15