448224 - fxdbug-win32-tbox has been intermittently orange since 2008-07-21

Status: RESOLVED WORKSFORME

(Core :: DOM: Core & HTML, defect)

Description

[Nick Thomas [:nthomas] (UTC+12)]

Attachment: Excerpt of tinderbox log

For somewhere between a quarter and a half of the recent builds, the Windows debug box for Firefox 3.0.x goes orange doing a trace-malloc test. See the attachment for details. 

There's a few things this could be 
* a problem with the tinderbox, given the mac and linux machines are OK.
I've clobbered the box just now and will check back after a few cycles to see if it's working
* there was a code change in (roughly) this window 
http://bonsai.mozilla.org/cvsquery.cgi?treeid=default&module=PhoenixTinderbox&branch=HEAD&branchtype=match&dir=&file=&filetype=match&who=&whotype=match&sortby=Date&hours=2&date=explicit&mindate=2008%2F07%2F20+00%3A00&maxdate=2008%2F07%2F21+16%3A00&cvsroot=%2Fcvsroot
* bug 448222, it seems the proper content on mozilla.org went away as part of the re-org. Doesn't really explain why there should sometimes be a leak on the "pretty" 404 page.

Comment 1

[Nick Thomas [:nthomas] (UTC+12)]

Reed confirmed in bug 448222 that there were changes on mozilla.org at 2008-07-20 00:29, so it's a good time-match for this issue appearing. At that point it started returning a pretty 404 page for 
  http://www.mozilla.org/newlayout/samples/test2.html
(This is loaded via bloatcycle.html). The previous content is now at 
 http://www-archive.mozilla.org/newlayout/samples/test2.html

Comment 2

[Nick Thomas [:nthomas] (UTC+12)]

Attachment: The page causing the intermittent crash

dbaron says this looks like a crash on a fatal assertion. We set XPCOM_DEBUG_BREAK=stack-and-abort for this tinderbox.

Comment 3

[Nick Thomas [:nthomas] (UTC+12)]

See also bug 431938.

Comment 4

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

Found in triage. confirm this is still intermittently orange, and then back to green again. 

Dealing with other releases right now, dont have time to deal with this just yet. Will come back for this once we come up for air.

Comment 5

[John O'Duinn [:joduinn] (please use "needinfo?" flag)]

Seems that bug#448222 causes a bloatcycle test to return 404 for some pages, which causes intermittent crashes. See bug#431938 for another possible example? 

I believe nothing here for RelEng to do, reassigning as best as I can guess.

Comment 6

[Nick Thomas [:nthomas] (UTC+12)]

The tests got restored in bug 451826, but this box is still doing a traffic light impression.

Comment 7

[Nick Thomas [:nthomas] (UTC+12)]

Attachment: 2nd excerpt from tinderbox log

Looks like it's an intermittent leak of 92 bytes in nsHostResolver now:

Error: Leak Test Failed: Number of leaks 92 is greater than LeakFailureThreshold 0

Full log:
http://tinderbox.mozilla.org/showlog.cgi?tree=Firefox3.0&errorparser=unix&logfile=1220486940.1220487297.27221.gz&buildtime=&buildname=Linux%20fxdbug-linux-tbox%20Depend&fulltext=1

Comment 8

[Nick Thomas [:nthomas] (UTC+12)]

Whoopsie, the RLk is on the linux box fxdbug-linux-tbox. The windows box might be alright now ?

Comment 10

[Nick Thomas [:nthomas] (UTC+12)]

It's green now, the reason for that is lost to the mysts of time.