1215686 - Disable ASAN leak reporting on reftests for now

Status: RESOLVED FIXED

(Testing :: Reftest, defect)

Comment 1

[(no longer active)]

Attachment: Disable ASAN leak reporting on reftests for now

Comment 2

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/7a8d20f52da6

Comment 3

[Phil Ringnalda (:philor)]

Fascinating state of affairs: doing that causes jsreftest to be permared, https://treeherder.mozilla.org/logviewer.html#?job_id=15801983&repo=mozilla-inbound, and I suspect will eventually prove to cause reftest-1 to be permared, https://treeherder.mozilla.org/logviewer.html#?job_id=15798381&repo=mozilla-inbound, though it will take a while to know about that because we can't be bothered to run ASan reftest-1 more than once every seven pushes, so a backfill is currently running.

Backed out in https://hg.mozilla.org/integration/mozilla-inbound/rev/7d62df943e5e

Comment 4

[(no longer active)]

Seems like we're running out of memory:

17:50:43     INFO -  REFTEST TEST-LOAD | file:///builds/slave/test/build/tests/reftest/tests/layout/reftests/bugs/632781-verybig.html | 4225 / 6448 (65%)
17:50:50     INFO -  ==1926==ERROR: AddressSanitizer failed to allocate 0x17d7b000 (400011264) bytes of LargeMmapAllocator: 12

But it's not clear what this means:

20:12:12  WARNING -  TEST-UNEXPECTED-FAIL | file:///builds/slave/test/build/tests/jsreftest/tests/jsreftest.html?test=test262/intl402/ch09/9.2/9.2.6_4_b.js | application terminated with exit code -9

Andrew, I think you should retry landing this together with the clang 3.7 upgrade when everything is ready.  What do you think?

Comment 5

[Andrew McCreight [:mccr8]]

The standard ASan environment settings reduce the amount of memory it uses, so my guess is that your patch is overwriting some existing environment variables. The right fix then would probably be to track down where it sets those in the first place. I can look into that.

Comment 6

[Andrew McCreight [:mccr8]]

Out of these places, it is probably automation.py.in

http://mxr.mozilla.org/mozilla-central/search?string=ASAN_OPTIONS&filter=[Aa]SAN_OPTIONS

Comment 7

[Andrew McCreight [:mccr8]]

Sorry, I should have realized that when I was reviewing.

Comment 8

[(no longer active)]

Ah makes sense!  Let me test a fix on the try server.

Comment 9

[(no longer active)]

Appending the environment variable fixes things, as expected: https://treeherder.mozilla.org/#/jobs?repo=try&revision=6c62f7dff307

Comment 10

[Pulsebot]

https://hg.mozilla.org/integration/mozilla-inbound/rev/926311d9197d

Comment 11

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/926311d9197d

Comment 12

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/926311d9197d