716869 - Some libxul addresses are never resolved in crash stacks/signatures

Status: RESOLVED INCOMPLETE

(Toolkit :: Crash Reporting, defect)

Description

[Robert Kaiser]

bug 711973 is filed on a skiplist addition in Socorro but the underlying problem is that a number of libxul.so/xul.dll/XUL addresses in stacks/signatures are not resolved to proper (function) names.

Judging from comments in e.g. bug 715916, but also general observation, this seems to be a regression, apparently somewhere between 11 and 12 (or probably somewhere in the 12 trunk timeframe).

This makes crash analysis pretty hard in some cases and has an especially large impact on native Android builds, it seems, so we should try to get this investigated and resolved ASAP.

Comment 1

[Scoobidiver (away)]

I think it's a dupe of bug 559661 that depends on bug 528092.

Comment 2

[Alex Keybl [:akeybl]]

(In reply to Robert Kaiser (:[e-mail]) from comment #0)
> This makes crash analysis pretty hard in some cases and has an especially
> large impact on native Android builds, it seems, so we should try to get
> this investigated and resolved ASAP.

Does this continue to affect crash analysis?

Comment 3

[Naoki Hirata :nhirata (please use needinfo instead of cc)]

Most of the libxul have to do with the arm_neon_fill/fast_composite_scaled_bilinear_neon_8888_8888_none_SRC ; however there are some crashes in 12 that might not be related to the arm_neon_fill crashes.

https://crash-stats.mozilla.com/report/index/1c2edee8-5b95-48d0-9d47-cf8b22120205
https://crash-stats.mozilla.com/report/index/d5806d8e-0038-4012-bd2f-a17dd2120205

https://crash-stats.mozilla.com/report/list?range_value=7&range_unit=days&date=2012-02-06&signature=libxul.so%400xaf1324%20|%20arm_neon_fill&version=FennecAndroid%3A13.0a1

https://crash-stats.mozilla.com/report/list?range_value=7&range_unit=days&date=2012-02-06&signature=libxul.so%400xafbfe8%20|%20fast_composite_scaled_bilinear_neon_8888_8888_none_SRC&version=FennecAndroid%3A13.0a1

Comment 4

[(not currently active) Ted Mielczarek]

Naoki: I think there's a big difference between what this was originally reported about, which AFAICT was "there are a few frames in a crash report with xul@address", vs. "there are no symbols for libxul at all in this report", which is what those two look like. If you're seeing a high incidence of crash reports like that, please file a new bug and we'll look into it. If it's just a few scattered reports then it might just be people using tinderbox builds or self builds that we don't have symbols for.

Comment 5

[Scoobidiver (away)]

On desktop, there are about 9,000 crash reports containing xul.dll or XUL across all versions:
https://crash-stats.mozilla.com/query/query?product=Firefox&version=ALL%3AALL&range_value=1&range_unit=weeks&query_search=signature&query_type=contains&query=xul.dll&reason=&build_id=&process_type=any&hang_type=any&do_query=1
https://crash-stats.mozilla.com/query/query?product=Firefox&version=ALL%3AALL&platform=mac&range_value=1&range_unit=weeks&query_search=signature&query_type=contains&query=XUL%40&reason=&build_id=&process_type=any&hang_type=any&do_query=1

On Fennec, there are about 300 crash reports containing libxul.so across all versions:
https://crash-stats.mozilla.com/query/query?product=Fennec&product=FennecAndroid&version=ALL%3AALL&range_value=1&range_unit=weeks&query_search=signature&query_type=contains&query=libxul.so&reason=&build_id=&process_type=any&hang_type=any&do_query=1

Comment 6

[Johnathan Nightingale [:johnath]]

Ted - based on comment 5, should this be tracked against ff12?

Comment 7

[(not currently active) Ted Mielczarek]

I suspect the xul.dll ones are related to bug 726570. I'm not sure about the non-Windows ones, but it's possible that they're just stackwalking issues as well. it's really hard to say without examining individual reports.

Comment 8

[(not currently active) Ted Mielczarek]

Looking at a few of the mobile crashes, it looks like it might be a similar issue. Most likely it's just bug 697301.

Comment 9

[Alex Keybl [:akeybl]]

(In reply to Robert Kaiser (:[e-mail]) from comment #0)
> This makes crash analysis pretty hard in some cases and has an especially
> large impact on native Android builds, it seems, so we should try to get
> this investigated and resolved ASAP.

If the impact is most noticeable for Native builds, we're not shipping them off of FF12 so no need to track for that release. Please re-nominate if this bug is getting in the way of desktop crash analysis.

Comment 10

[Benjamin Smedberg]

This bug as filed isn't really fixable or useful. If you have particular crash reports or patterns of crash reports which are showing incorrect results, please file them specifically. Note that a set of Windows stackwalking issues were fixed in production yesterday.