Closed Bug 463263 Opened 16 years ago Closed 16 years ago

leaking nsHostResolver a lot on multiple platforms from unrelated checkins

Categories

(Core :: General, defect)

defect
Not set
blocker

Tracking

()

RESOLVED FIXED
mozilla1.9.1b2

People

(Reporter: dietrich, Assigned: mcmanus)

References

Details

requesting blocking due to the amount of noisy orange created on the tree. assigning to patrick per bz's request.
Flags: blocking1.9.1?
Dietrich, I'm not that familiar with all the lingo. Can you point me to the test/description/data/recreate for this so I can address your specific problem?
Hi Patrick, the problem is visible on the tinderbox:

http://tinderbox.mozilla.org/showbuilds.cgi?tree=Firefox

The orange builds between 8 and 10am PST are largely due to leaks, as shown in this log:

http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1225911799.1225912078.20218.gz

(search for nsHostResolver)

The checkins at the time are not likely to have caused the leak. Boris said you knew about the problem, and could address it.
--ALL-LEAKS-----------------------------------leaks------leaks%-----------------------
TOTAL                                           368       0.00%
nsHostResolver                                  140       0.00%
nsDNSAsyncRequest                                96       0.00%
nsHostRecord                                     60       0.00%
nsHTMLDNSPrefetch                                56       0.00%
nsStringBuffer                                   16       0.00%
There is some info on leak test tools here:
https://developer.mozilla.org/en/Debugging_memory_leaks
and sparser info here:
https://wiki.mozilla.org/Performance:Leak_Tools

I find that looking at the commands in the tinderbox log gives the best documentation on which commands are useful.
Thanks for the background karl and dietrich.. I'll look into it now. The problem is no doubt with code I wrote that impacts those classes and was just checked in, but I don't at this time know specifically what is causing the problem.
Patrick: Which bug/checkin caused this? If there's not a obvious cause (and thus fix), we should back it out while it's investigated.
I'm assuming one of these was the cause:

19b3caf108d1	Boris Zbarsky — Bug 453403. Fix leak of nsHostResolver by making
                  sure to wake up all the worker threads on shutdown
e8fd3f4c52b6	Patrick McManus — Bug 453403. Add DNS prefetching, similar to
                  what Google chrome does. r+sr=bzbarsky
39fd2b88d622	Patrick McManus — Bug 459724. Make page reload bypass the DNS
                  cache. r+sr=bzbarsky
4542793088dc	Patrick McManus — Bug 459724. Fix normalization of flags in the
                  DNS cache to allow other work that depends on these flags to
                  happen. r+sr=bzbarsky

I'm going to back them all out.
justin - no doubt contained in he first two (and they should be applied and removed together).
Ok, backed out 19b3caf108d1 and e8fd3f4c52b6 (for bug 453403). As long as things stay green, the other two changesets (for bug 459724) can stay.
This seems to have fixed the problem at hand.
Status: NEW → RESOLVED
Closed: 16 years ago
Resolution: --- → FIXED
> See also bug 462410.

That doesn't seem related (in particular, it predates these leak fixes).
Blocks: 453403
Flags: blocking1.9.1?
You need to log in before you can comment on or make changes to this bug.