Closed
Bug 1348101
Opened 8 years ago
Closed 8 years ago
tc win64 tests use win32-minidump_stackwalk.exe
Categories
(Testing :: General, enhancement)
Testing
General
Tracking
(Not tracked)
RESOLVED
INVALID
People
(Reporter: gbrown, Assigned: gbrown)
Details
Several current win64 Talos bugs have timeouts or crashes that fail to produce crash reports. I noticed that we are calling win32-minidump_stackwalk.exe in those jobs. Hopefully a 64 bit variant will work better.
![]() |
Assignee | |
Comment 1•8 years ago
|
||
![]() |
Assignee | |
Comment 2•8 years ago
|
||
All the win64 tests pull win32-minidump_stackwalk.exe from tooltool.
Summary: Talos win64 uses win32-minidump_stackwalk.exe → tc win64 tests use win32-minidump_stackwalk.exe
![]() |
Assignee | |
Comment 3•8 years ago
|
||
:ted - Do we need a 64-bit variant of minidump_stackwalk for win64; if so, where would that come from?
Flags: needinfo?(ted)
Comment 4•8 years ago
|
||
I verified in mozinfo.info that we have:
windows8:
bits: 64
processor: x86_64
windows7:
bits: 32
processor: x86
this is both for mochitest and talos, they are identical.
looking in taskcluster this is hardcoded:
https://dxr.mozilla.org/mozilla-central/source/testing/mozharness/configs/unittests/win_unittest.py#276
and on linux it is conditional:
https://dxr.mozilla.org/mozilla-central/source/testing/mozharness/configs/talos/linux_config.py#8
but I noticed we do not have a tooltool entry for win64 minidump:
https://dxr.mozilla.org/mozilla-central/search?q=path%3Atooltool-manifests&redirect=true (at the bottom is testing/../win32/releng.manifest)
so I think the next step is to get a win64 stackwalk_minidump on tooltool- possibly we wait for ted to weigh in before hacking too much in case there is a reason for this?
Comment 5•8 years ago
|
||
(In reply to Geoff Brown [:gbrown] from comment #3)
> :ted - Do we need a 64-bit variant of minidump_stackwalk for win64; if so,
> where would that come from?
There should be no real reason to need a 64-bit minidump_stackwalk unless the 32-bit binary is OOMing or something silly like that. minidump_stackwalk is written specifically to handle minidumps from any platform (we process minidumps from Firefox on every OS/architecture on 64-bit Linux on crash-stats).
FWIW, those binaries were built in taskcluster using this script:
https://dxr.mozilla.org/mozilla-central/source/taskcluster/scripts/misc/minidump_stackwalk.sh
All the binaries are cross-compiled from a single Linux taskcluster task, since a) we didn't have any non-Linux workers at the time and b) it's probably easier to wrangle the cross-compilation in this case than to deal with the headaches of native Windows/OS X compilation. If we really did need a win64 binary we'd have to update the Docker image to have the right 64-bit mingw toolchain and tweak the configure targets appropriately.
We could create a toolchain task to run that script nowadays, that would be handy for generating new binaries.
Flags: needinfo?(ted)
Comment 6•8 years ago
|
||
So as a first cut--do we get minidump files out of those Talos crashes at all? Even if they fail to produce a stack? For unittest jobs we upload the minidumps as test artifacts, do we do that for Talos? (If not we should fix that.) If so, we should be able to look at the minidumps and see if they have useful data first.
![]() |
Assignee | |
Comment 7•8 years ago
|
||
Thanks Ted. I expect there's nothing else to do here then.
We usually get minidump files from those Windows 8 Talos crashes, and they appear as test artifacts, but minidump_stackwalk fails. Bug 1310638 (which you already found and commented on today) has lots of examples.
Assignee: nobody → gbrown
Status: NEW → RESOLVED
Closed: 8 years ago
Resolution: --- → INVALID
You need to log in
before you can comment on or make changes to this bug.
Description
•