Closed Bug 328219 Opened 20 years ago Closed 20 years ago

crash when reading an rss feed because gIOService is null

Categories

(MailNews Core :: Feed Reader, defect)

x86
Windows XP
defect
Not set
critical

Tracking

(Not tracked)

RESOLVED FIXED

People

(Reporter: Bienvenu, Assigned: Bienvenu)

Details

(4 keywords)

Attachments

(1 file)

Some http changes have exposed a bug in Feed.js - we're creating an instance of the io service instead of calling getService. I'll attach a fix.
Attached patch proposed fixSplinter Review
Attachment #212785 - Flags: superreview?(mscott)
Attachment #212785 - Flags: superreview?(mscott) → superreview+
Status: NEW → RESOLVED
Closed: 20 years ago
Resolution: --- → FIXED
fixed on 1.8.1 branch as well, since I think the http changes landed there as well.
Keywords: fixed1.8.1
So I've been using Thunderbird 1.5.0.* as an RSS reader, and leaving it up for many days on end. I've been noticing for quite a while that when I quit after using Thunderbird for a long time, it pegs the CPU for a while, sometimes even a few minutes. So today I broke into the pegged CPU in the debugger (in the release build) and from the few symbols and the disassembly, determined that it was in a tight loop in nsPrefBranch::freeObserverList, called from the "xpcom-shutdown" observer in NS_ShutdownXPCOM. So I did a build of the 1.8.0 branch, put a printf in freeObserverList, and found that the observers being removed were for the prefs "network.security.ports." and "network.autodial-helper.enabled". Since the IO service is the only thing that registers observers for these prefs, I concluded that it was being created multiple times (somebody was using createInstance rather than getService), but a breakpoint in its constructor, and found that the guilty party was the code fixed in this patch. Then, after noticing that the problem was fixed on trunk and 1.8 branch, I tied it back to this bug. So I think it would be good to get this patch in on the 1.8.0 branch to fix the leak of pref observers and hang on shutdown (and presumably hang if either of those prefs ever changes).
Flags: wanted1.8.0.x?
Keywords: mlk
I agree, this would be a very good bug to take on the branch. Thanks to dbaron for tracking down the hang he was seeing and figuring out that it was this bug.
Attachment #212785 - Flags: approval1.8.0.10?
Comment on attachment 212785 [details] [diff] [review] proposed fix approved for 1.8.0 branch, a=dveditz for drivers
Attachment #212785 - Flags: approval1.8.0.10? → approval1.8.0.10+
fixed on 1.8.0.x branch
Keywords: fixed1.8.0.10
Severity: normal → critical
Flags: wanted1.8.0.x?
Keywords: crash
Could someone suggest how to test this?
Comment on attachment 212785 [details] [diff] [review] proposed fix 1. get the ioservice 2. use the subscriptloader to load the chrome url for: mozilla/mail/extensions/newsblog/content/Feed.js 3. FeedCache.normalizeHost(null) 4. quit() at least, theoeritcally
Component: RSS → Feed Reader
Product: Thunderbird → MailNews Core
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Created:
Updated:
Size: