WebKit Bugzilla
New
Browse
Log In
×
Sign in with GitHub
or
Remember my login
Create Account
·
Forgot Password
Forgotten password account recovery
RESOLVED FIXED
159807
js/stringimpl-to-jsstring-on-large-strings tests consume huge memory
https://bugs.webkit.org/show_bug.cgi?id=159807
Summary
js/stringimpl-to-jsstring-on-large-strings tests consume huge memory
Csaba Osztrogonác
Reported
2016-07-15 04:08:47 PDT
https://bugs.webkit.org/show_bug.cgi?id=158793#c6
(In reply to
comment #4
)
> Comment on
attachment 281956
[details]
> Patch > > Clearing flags on attachment: 281956 > > Committed
r202415
: <
http://trac.webkit.org/changeset/202415
>
These new tests consume huge memory, my PC with 8Gb memory started to swap and became unresponsible for long time when I exectuted run-javascriptcore-tests. :( It would be good not adding tests need 16-32-64Gb memory to be able to run. :(
Attachments
Patch
(5.86 KB, patch)
2016-09-14 04:01 PDT
,
Csaba Osztrogonác
saam
: review+
saam
: commit-queue-
Details
Formatted Diff
Diff
View All
Add attachment
proposed patch, testcase, etc.
Alexey Proskuryakov
Comment 1
2016-07-15 20:45:09 PDT
> when I exectuted run-javascriptcore-tests. :(
This should affect run-webkit-tests too, correct?
Saam Barati
Comment 2
2016-07-16 22:21:31 PDT
(In reply to
comment #0
)
>
https://bugs.webkit.org/show_bug.cgi?id=158793#c6
> > (In reply to
comment #4
) > > Comment on
attachment 281956
[details]
> > Patch > > > > Clearing flags on attachment: 281956 > > > > Committed
r202415
: <
http://trac.webkit.org/changeset/202415
> > > These new tests consume huge memory, my PC with 8Gb memory > started to swap and became unresponsible for long time > when I exectuted run-javascriptcore-tests. :( > > It would be good not adding tests need 16-32-64Gb memory to be able to run. > :(
I agree. Ben, is there a way to write a similar test for the original bug without consuming as much memory?
Csaba Osztrogonác
Comment 3
2016-07-18 05:35:29 PDT
(In reply to
comment #1
)
> > when I exectuted run-javascriptcore-tests. :( > > This should affect run-webkit-tests too, correct?
It could, but the problem is bigger with run-javascriptcore-tests, because it runs each test with different configurations parallel, for example: Running jsc-layout-tests.yaml/js/script-tests/stringimpl-to-jsstring-on-large-strings-2.js.layout Running jsc-layout-tests.yaml/js/script-tests/stringimpl-to-jsstring-on-large-strings-2.js.layout-no-llint Running jsc-layout-tests.yaml/js/script-tests/stringimpl-to-jsstring-on-large-strings-2.js.layout-no-cjit Running jsc-layout-tests.yaml/js/script-tests/stringimpl-to-jsstring-on-large-strings-2.js.layout-dfg-eager-no-cjit It means that 4 copy of the same memory consuming test can run in the same time.
Benjamin Poulain
Comment 4
2016-07-19 14:10:26 PDT
The problem is this bug only crashed JSC when dealing with strings larger than 2gb. We need bugs like this to be covered but we can't have these tests break other tests. Ossy, is there any way to make some tests sequential instead of running in parallel with everything else? Any other idea to solve this?
Csaba Osztrogonác
Comment 5
2016-07-20 07:44:38 PDT
(In reply to
comment #4
)
> The problem is this bug only crashed JSC when dealing with strings larger > than 2gb. > > We need bugs like this to be covered but we can't have these tests break > other tests. > > Ossy, is there any way to make some tests sequential instead of running in > parallel with everything else? Any other idea to solve this?
run-jsc-stress-tests uses GNU make (-jXXX) to run tests parallel, there is no way to prevent running them parallel without giant hacks in prepareMakeTestRunner:
https://trac.webkit.org/browser/trunk/Tools/Scripts/run-jsc-stress-tests#L1697
Do we really need to run these test in all configurations? - default - no-llint - no-cjit - dfg-eager-no-cjit I suggest skipping them until we find a better fix.
Csaba Osztrogonác
Comment 6
2016-09-14 04:01:37 PDT
Created
attachment 288803
[details]
Patch
Saam Barati
Comment 7
2016-09-14 16:52:40 PDT
Comment on
attachment 288803
[details]
Patch View in context:
https://bugs.webkit.org/attachment.cgi?id=288803&action=review
r=me
> Tools/Scripts/run-javascriptcore-tests:137 > + Skip tests tagged with //\@large-heap
@largeHeap not @large-heap
Csaba Osztrogonác
Comment 8
2016-09-15 04:15:43 PDT
(In reply to
comment #7
)
> Comment on
attachment 288803
[details]
> Patch > > View in context: >
https://bugs.webkit.org/attachment.cgi?id=288803&action=review
> > r=me > > > Tools/Scripts/run-javascriptcore-tests:137 > > + Skip tests tagged with //\@large-heap > > @largeHeap not @large-heap
Good catch, I copied this typo from run-jsc-stress-tests script. I'll fix there too before landing.
Csaba Osztrogonác
Comment 9
2016-09-15 04:20:22 PDT
Committed
r205967
: <
http://trac.webkit.org/changeset/205967
>
Note
You need to
log in
before you can comment on or make changes to this bug.
Top of Page
Format For Printing
XML
Clone This Bug