LGTM on the limit in d8.cc, just a reliable test case is needed.
https://codereview.chromium.org/1284683004/diff/1/test/mjsunit/regress/regress-crbug-518748.js
File test/mjsunit/regress/regress-crbug-518748.js (right):
https://codereview.chromium.org/1284683004/diff/1/test/mjsunit/regress/regress-crbug-518748.js#newcode8
test/mjsunit/regress/regress-crbug-518748.js:8: for (var i = 0; new
Worker('onmessage = function() {};'); i++) a[i] = i;
On 2015/08/10 22:09:47, binji wrote:
On 2015/08/10 at 22:01:23, Michael Starzinger wrote:
> The regression test is kind of brittle, because if the workers
finish fast
enough, there will never be more than 100 of them running in parallel.
We need
some more reliable way to test that. Also a loop that doesn't run
infinitely but
is bounded somehow would be better.
Workers are not reaped until the main thread is finished, so this
should not be
a problem. But if someone decided to fix that, you're right. In any
case, there
needs to be a better test because this is failing on multiple bots
as-is. :)
Ah, got it. My bad, I was assuming the {workers} list in d8.cc would be
emptied out whenever a worker terminated. But yes, as you mentioned, for
some reason some try-bots still fail on this test, so it's brittle for
other reasons.
https://codereview.chromium.org/1284683004/
--
--
v8-dev mailing list
[email protected]
http://groups.google.com/group/v8-dev
---
You received this message because you are subscribed to the Google Groups "v8-dev" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to [email protected].
For more options, visit https://groups.google.com/d/optout.
