Closed Bug 572190 Opened 14 years ago Closed 14 years ago

"Parsing tinderbox failed" when going back in time on MozillaTry

Categories

(Tree Management Graveyard :: TBPL, defect)

x86
Linux
defect
Not set
normal

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: dbaron, Unassigned)

References

(Blocks 1 open bug, )

Details

Attachments

(2 files)

Currently (and quite frequently, I think; I've seen this before), when I load http://tests.themasta.com/tinderboxpushlog/?tree=MozillaTry and click the green down arrow at the bottom to go back 12 hours, I get "Parsing Tinderbox failed. :-(" at the top, and no build data except for the top few pushes (whose builds happened in the following 12 hours).

This means that if I want to get my try server results that weren't yet ready by the time I went to sleep last night, I need to push the same thing to try again just because I can't find the results.
Hmmm.  Interestingly enough, this actually works fine in Chrome, but I get the error in Firefox (mozilla-central).
As far as I can tell (on Linux x86), we're failing to create a parse node (due to an arena allocation returning null) parsing the not-quite-json that tinderbox spits out (I was looking at http://tinderbox.mozilla.org/Firefox/json.js?_=1276647465822 - 4.8 megs in, I don't know how many megs of parse nodes we create before giving up).
Actually, I meant x86-64.
I'm not claiming this is a complete fix here or anything, but it does make it so that the JS stack limits we set up are a function of the size of size_t, which should make the difference between 32-bit and 64-bit systems much much smaller.
Attachment #457690 - Flags: review?(mrbkap)
Attachment #457690 - Flags: review?(mrbkap) → review+
Attachment 457690 [details] [diff] checked in as http://hg.mozilla.org/mozilla-central/rev/b9f320e24e04. Leaving this bug open since there's probably more to do here.
jst, mind opening a second bug? bugzilla bugs with multiple checked-in patches tend to be confusing.
This makes the test pass if either 'InternalError: script stack space quota is exhausted' or 'InternalError: allocation size overflow' is thrown in js1_5/extensions/regress-336{409,410}.js
Attachment #458947 - Flags: review?(jst)
Comment on attachment 458947 [details] [diff] [review]
patch to js1_5/extensions/regress-336{409,410}.js

+var expect = /(No Crash|InternalError: script stack space quota is exhausted|nternalError: allocation size overflow)/;

r=jst if you add the missing 'I' after the second '|', as you pointed out on irc :)
Attachment #458947 - Flags: review?(jst) → review+
I think this might have caused tinderbox to go orange on 64bit systems
The test changes weren't good enough. After discussing this with Bob we decided to disable the two offending tests when run in the browser on a 64-bit system. Pushed as:

http://hg.mozilla.org/mozilla-central/rev/1ac07fe5f6c9
I'm still seeing this stack-space exhaustion from TBPL currently.
Blocks: 583098
Yup, this is still not fully fixed. Since I hijacked this bug for the partial fix (stack size on 32 vs 64 bit systems) I just cloned this bug as bug 583098 and I'm marking this one fixed. Please keep any followup that's not specific to the stack size changes that were made here in bug 583098, or other bugs dependent on it.
No longer blocks: 583098
Status: NEW → RESOLVED
Closed: 14 years ago
Resolution: --- → FIXED
Blocks: 583098
Product: Webtools → Tree Management
Product: Tree Management → Tree Management Graveyard
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: