Closed Bug 722044 Opened 13 years ago Closed 13 years ago

[skiplist] - please add TouchBadMemory to prefixSignatureRegEx

Categories

(Socorro :: Infra, task)

task
Not set
major

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: MatsPalmgren_bugz, Assigned: laura)

References

Details

(Whiteboard: [qa-])

Please add TouchBadMemory to skiplist. Example crash report: bp-1a11b3bd-a0ff-4423-993f-674582120121 The relevant stack frame there is #3.
It's #2 top crasher in FennecAndroid 12.0a1 with a bunch of different stacks so it will help to breakdown.
Severity: normal → major
Component: General → Infra
QA Contact: general → infra
Laura, can you take this?
Blocks: 721667
Blocks: 705641
Blocks: 727140
Remove bug 727410 from dep change, it was resolved without this, but this is an important fix still.
No longer blocks: 727140
Blocks: 589546
No longer blocks: 721342
No longer blocks: 717663, 720092
Blocks: 731296
Blocks: 733322
Blocks: 733323
Blocks: 733324
Blocks: 733325
Please also add MOZ_Crash and MOZ_Assert (example: bp-1eb2b5ed-7e75-4c76-bf00-255842120309)
Fwiw, I think how we currently handle this list is ridiculous. We need to get this list checked in to hg.m.o so that developers can add words to it as we see fit.
(In reply to Mats Palmgren [:mats] from comment #4) > Please also add MOZ_Crash and MOZ_Assert It's bug 733235.
(In reply to Mats Palmgren [:mats] from comment #5) > Fwiw, I think how we currently handle this list is ridiculous. > We need to get this list checked in to hg.m.o so that developers can add > words to it > as we see fit. This is a Socorro configuration and doesn't have a bit to do with hg (Socorro is on github). It's in the Socorro repo of course and you can always send a pull request for it if you like.
What I mean is, the skiplist should not be owned by the Socorro project. It should be owned by the users of https://crash-stats.mozilla.com/ ie. the developers of the products that reports its crashes there. It's we that have an interest in a good skiplist and we are the ones that knows what should be on it for it to be good. Having to wait around months for someone else to add a single word to a file and check it in is extremely frustrating. This doesn't work! Let's put this file in mozilla-central and managed by #developers, have a bot upload it and install it on crash-stats.mozilla.com on daily basis, or something... If you want it in github, fine, where do I apply for commit access? (I specifically do NOT want any Socorro people be the gatekeepers to changing the skiplist, they have proved they aren't good owners)
(In reply to Mats Palmgren [:mats] from comment #8) > What I mean is, the skiplist should not be owned by the Socorro project. I disagree. It only affects crash-stats.m.c so it should be somewhere closely connected to it, either in its site-specific configuration (where it is right now) or in the admin interface for it (bugs are filed for that, code contributions welcome). > Having to wait around months for someone else to add a single word > to a file and check it in is extremely frustrating. This doesn't work! It should never take months, I'm with you here. We have Socorro updates ship every week, things should always go into the next update. Sometimes the bugs are not written in an actionable way for the team, though. > Let's put this file in mozilla-central and managed by #developers No, that would make signatures change every day at random because someone would find out that something changes, and that would make tracking crashes over time completely useless (it's enough hassle to pull strings together when changing the skiplist now every week with Socorro releases). It also sometimes needs discussion of what list something should go on, etc. > If you want it in github, fine, where do I apply for commit access? You don't need direct commit access to the repo, if you send a pull request, people get notified and pull it in within a day, usually. > (I specifically do NOT want any Socorro people be the gatekeepers to > changing the skiplist, they have proved they aren't good owners) Now you are insulting Mozilla community members, that won't be tolerated.
fyi, mats, even if it was in hg; laura handles the requests pretty fast esp if you give her the regex correctly. I've had no issues with getting my requests within 24 hrs. The change however won't show until the next release of socorro if I'm not mistaken? The data doesn't get back processed, so you will still see a lingering of the old signature for a while.
Let's go to the heart of the matter, should this be in the prefixlist or in the ignorelist?
(In reply to Robert Kaiser (:kairo@mozilla.com) from comment #11) > Let's go to the heart of the matter, should this be in the prefixlist or in > the ignorelist? libmozalloc symbols are in the prefix skip list, which is the default one.
Blocks: 736220
Blocks: 736522
Blocks: 736526
Blocks: 736714
Blocks: 720353
blassey said on IRC that "TouchBadMemory() should be treated like abort()" and given that https://github.com/mozilla/socorro/blob/master/scripts/config/processorconfig.py.dist#L206 seems to be where abort() matches ans taking to account what comment #12 says, I guess we can be pretty sure that prefixlist is what we want here.
Summary: [skiplist] - please add TouchBadMemory to skiplist → [skiplist] - please add TouchBadMemory to prefixSignatureRegEx
Target Milestone: --- → 2.5.2
Whiteboard: [qa-]
Commits pushed to master at https://github.com/mozilla/socorro https://github.com/mozilla/socorro/commit/2bc0f372fa913d4f39a229cdd77eefa694a5e931 Fixes bug 722044, add TouchBadMemory to the prefix list https://github.com/mozilla/socorro/commit/07a9b519f4584464dc7c8d203a89898e1988bc7a Merge pull request #445 from lauraxt/722044 Fixes bug 722044, add TouchBadMemory to the prefix list
Status: NEW → RESOLVED
Closed: 13 years ago
Resolution: --- → FIXED
Adrian: can you cherry pick this for 2.5.2 please? It landed on master here: https://github.com/mozilla/socorro/commit/07a9b519f4584464dc7c8d203a89898e1988bc7a
Assignee: nobody → laura
You need to log in before you can comment on or make changes to this bug.