> We inject Ahem directly to content_shell, which is supposed to be reliable.
I agree with this logic, but it does not work in reality. See issue 392046 . That is why we started using ahem.js.
This bug is now blocking CQ for many CLs. Need to retry multiple times to pass CQ.
I created a CL https://chromium-review.googlesource.com/c/chromium/src/+/940637 before seeing this bug. I wonder if <meta name="flags" content="ahem"> would work to deflake ahem font loading.
The CL didn't work. Abandoned.
win7_chromium_rel_ng is very flaky because of the bug. Many blink CLs fail to CQ after several hours trying on the bot.
Should we revert the CLs which added the directories?
The issue seems fixed so far, the last builds have been successful on the bots.
It seems the same problem is affecting other tests too, I'm doing a similar CL to see if it fixes issue #794136 for example.
Comment 1 by r...@igalia.com
, Feb 27 2018