1692309 - Surface the GetLastError() value in Windows minidumps

Status: RESOLVED FIXED

(Toolkit :: Crash Reporting, enhancement)

Description

[Gabriele Svelto [:gsvelto]]

Attachment: Contents of GetLastError() as extracted via windbg

:dmajor pointed out in bug 1688271 comment 5 that Windows minidumps taken from the nightly contain both the value that GetLastError() would return at the point the minidump was taken. When analyzing a minidump with windbg this value can be extracted via the !gle command and it gets printed out both as a Win32 error and as the corresponding NTSTATUS value (IIUC). See the screenshot.

The value is stored in the TEB (bug 1688271 comment 7) whose address is stored in the minidump's thread information. We should modify the stack walker to extract it and print it out together with the rest of the stack-trace.

I'll file a separate bug to display it in Socorro's front-end.

Comment 1

[Will Kahn-Greene [:willkg] ET needinfo? me]

I started tracking stackwalker issues here: https://github.com/mozilla-services/minidump-stackwalk/issues

Seems like this is specifically a stackwalker issue. Does that sound right?

We can keep this bug. I can write an issue in the minidump-stackwalk repo pointing here for details.

Comment 2

[(Away)]

If the code ends up hardcoding the field offset (since the SDK's struct _TEB is opaque), remember that it's architecture-dependent. 0x34 on 32-bit and 0x68 on 64-bit.

Comment 3

[Aria Beingessner [:Gankra]]

Attachment: Bug 1692309 - extract the GetLastError value from minidumps. r?gsvelto

Comment 4

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/70904d6ca415
extract the GetLastError value from minidumps. r=gsvelto

Comment 5

[Alexandru Michis [:malexandru]]

https://hg.mozilla.org/mozilla-central/rev/70904d6ca415

Comment 6

[Will Kahn-Greene [:willkg] ET needinfo? me]

I don't understand how this bug is a bug in the Socorro processor. Should it get moved somewhere else? Is there work we have to do in the Socorro processor?

Comment 7

[Gabriele Svelto [:gsvelto]]

I messed things up. This should be in crash reporting (gotta move it there) and I should file another one for the processor which will be about including this feature in Socorro's stackwalker.

Comment 8

[Gabriele Svelto [:gsvelto]]

Actually, I'm going to put the new bug directly on Socorro's stackwalker GitHub project because it makes more sense there.

Comment 9

[Will Kahn-Greene [:willkg] ET needinfo? me]

Awesome--thank you!