Closed Bug 676499 Opened 13 years ago Closed 4 years ago

Mac 64 bit crash [@ fts3SegReaderNext] in Thunderbird's gloda tests following change to optimisation levels in sqlite

Categories

(Toolkit :: Storage, defect, P5)

x86_64
macOS
defect

Tracking

()

RESOLVED WORKSFORME

People

(Reporter: standard8, Unassigned)

References

Details

(Keywords: crash, regression)

Crash Data

Attachments

(2 files)

Attached file Test Logs
When running Thunderbird's gloda unit tests on an optimized build, we're seeing crashes @fts3SegReaderNext (see attached info).

This is occurring since bug 654049 landed on mozilla-central and is causing permanent orange.

Backing out just the sqlite3 part of http://hg.mozilla.org/mozilla-central/rev/86e9d6953e50 fixes the issue.

What do we think about backing out just that part until we can resolve the crash we're seeing? (I had a brief look, but wouldn't really know where to start in that code).
Just make it OSX-only.
Attached patch The fixSplinter Review
Here's the Mac-only exclusion which passes on my machine. I've also added a comment as to why we're forcing -O2, I can either reference this bug, or a new bug which I can file for the actual crash.
Assignee: nobody → mbanner
Status: NEW → ASSIGNED
Attachment #550667 - Flags: review?(mh+mozilla)
Comment on attachment 550667 [details] [diff] [review]
The fix

Review of attachment 550667 [details] [diff] [review]:
-----------------------------------------------------------------

I think using this bug is good enough. If you file a new bug for the underlying issue, just add a relationship with this one. I'm afraid there's not much to do about it, though, it's likely a gcc miscompilation with -O3, and Xcode using an old gcc...
Attachment #550667 - Flags: review?(mh+mozilla) → review+
Checked in: http://hg.mozilla.org/mozilla-central/rev/c7ea065539d2
Assignee: mbanner → nobody
Severity: blocker → critical
Status: ASSIGNED → NEW
(In reply to Mike Hommey [:glandium] from comment #3)
> not much to do about it, though, it's likely a gcc miscompilation with -O3,
> and Xcode using an old gcc...

Is this still true with Clang?
(In reply to Nomis101 from comment #5)
> (In reply to Mike Hommey [:glandium] from comment #3)
> > not much to do about it, though, it's likely a gcc miscompilation with -O3,
> > and Xcode using an old gcc...
> 
> Is this still true with Clang?

Nomis101, Are you asking because of crashes in releases?

FWIW, no Mac crashes on crash-stats for current versions in past 5 months - all are version 3.x.

Not so however for windows. For example bp-d62526ad-e19f-42e6-8b6e-41db12130124 is a user who had hundreds of crashes (but most were [@ memcpy | fts3SegWriterAdd] - per https://getsatisfaction.com/mozilla_messaging/topics/thunderbird_17_0_2_continually_crashes_about_5_seconds_after_startup
Flags: needinfo?(mh+mozilla)
(In reply to Wayne Mery (:wsmwk) from comment #6)
> (In reply to Nomis101 from comment #5)
> > (In reply to Mike Hommey [:glandium] from comment #3)
> > > not much to do about it, though, it's likely a gcc miscompilation with -O3,
> > > and Xcode using an old gcc...
> > 
> > Is this still true with Clang?
> 
> Nomis101, Are you asking because of crashes in releases?

No, I'm not asking about crashes. I'm asking because this patch was because of (likely) a gcc miscompilation with -O3. But now we are not using gcc anymore, we are using Clang now. And my question was, if this bug (crashes) will still occur if sqlite gets compiled with -O3 and Clang.
We could try.
Flags: needinfo?(mh+mozilla)
(In reply to Mike Hommey [:glandium] from comment #8)
> We could try.

FWIW, there are almost no crashes in current versions of 52.x
maybe we should resolve as inactive?
Priority: -- → P5

Closing because no crashes reported for 12 weeks.

Status: NEW → RESOLVED
Closed: 4 years ago
Resolution: --- → WORKSFORME
You need to log in before you can comment on or make changes to this bug.