1367103 - [meta] Fix TSAN issues in the GC

Status: NEW

(Core :: JavaScript: GC, task, P3)

Comment 1

[Yoshi Cheng-Hao Huang [:allstars.chh][:allstarschh][:yoshi]]

Hi Jonco
I'd like to help on these bugs, but I am not sure how to reproduce these bugs.

Recently I fixed build error for TSan in bug 1404839, however once I started Firefox, it also showed lots of TSan error came from many componets of Gecko but they are not related to JS Engine.

I also looked into these bugs in this meta bug, but the first comment is always the TSan log, so I still don't know how to reproduce it, or how to verify my patch with Tsan.

Can you give me some suggestions about this?

thanks

Comment 2

[Jon Coppeard (:jonco)]

(In reply to Yoshi Huang [:allstars.chh] from comment #1)
Thanks for looking at this.  This is an important area where we have a lot of work to do.

I don't know how these bugs where originally found.  You could needinfo the reporter.  A lot of them are quite old though, and I suspect that they might not show up in the same way today.

One thing you could do would be to be build the shell with TSAN enabled and investigate any issues you find when running the tests.  I expect there will be a few, which we may or may not have open bugs for.

The plan for TSAN is to have the tests run as part of the normal builds (bug 1291954) so that we have a simple way of seeing what the current problems are and verifying when we have fixed them.

Comment 3

[Yoshi Cheng-Hao Huang [:allstars.chh][:allstarschh][:yoshi]]

Nathan, I found most bugs are filed by you, but do you still have the test script to trigger the failure? 
Because in these bugs there are only stack traces, however those bugs are more than 2 years, the stack might be changed a lot.

If you still have the test script, I could help to see if the problems still exist in current m-c, and help to fix them. :)

Thanks

Comment 4

[Nathan Froyd [:froydnj]]

(In reply to Yoshi Huang [:allstars.chh] from comment #3)
> Nathan, I found most bugs are filed by you, but do you still have the test
> script to trigger the failure? 
> Because in these bugs there are only stack traces, however those bugs are
> more than 2 years, the stack might be changed a lot.
> 
> If you still have the test script, I could help to see if the problems still
> exist in current m-c, and help to fix them. :)

That'd be great!  The test script wasn't any more sophisticated than running tests (usually just a single directory's worth of mochitests or xpcshell tests) and filing the relevant stack traces as bugs.  As bugs got fixed, I would run a different subset of tests, and that would often turn up more bugs.