Closed Bug 212598 Opened 21 years ago Closed 21 years ago

netwerk/ should not use nsIPref

Categories

(Core :: Networking, defect)

defect
Not set
normal

Tracking

()

RESOLVED FIXED

People

(Reporter: darin.moz, Assigned: darin.moz)

References

Details

Attachments

(1 file, 1 obsolete file)

netwerk/ should not use the deprecated nsIPref. this blocks bug 175193.
Attached patch v1 patch (obsolete) — Splinter Review
Attachment #127678 - Flags: review?(dougt)
Attachment #127678 - Flags: superreview?(bzbarsky)
Comment on attachment 127678 [details] [diff] [review] v1 patch what part about this isn't thread safe? + // XXX not threadsafe I wonder what would happen if we ever had unsolicited notifications: +nsProtocolProxyService::Observe(nsISupports *aSubject, + const char *aTopic, + const PRUnichar *aData) +{ + PrefsChanged(aTopic); assuming it works... r=dougt
Attachment #127678 - Flags: review?(dougt) → review+
doug: operator++ is not threadsafe. we'd need to use PR_AtomicIncrement to make that code threadsafe. i don't think it currently matters any since the dir index parser is only run on the UI thread, but the NS_IMPL_THREADSAFE_xxx macro seems to suggest otherwise. can there be unsolicited notifications? that would just be a major bug in the observer service, no?
Attached patch v1.1 patchSplinter Review
revised per comments from caillon over IRC: 1- use NS_PREFSERVICE_CONTRACTID instead of NS_PREFSERVICE_CID 2- make ns{FTP,Gopher}Channel::SetListFormat reject invalid values in prefs
Attachment #127678 - Attachment is obsolete: true
Attachment #127678 - Flags: superreview?(bzbarsky)
Comment on attachment 127689 [details] [diff] [review] v1.1 patch minor changes over v1 patch... please see previous comments in bug for details.
Attachment #127689 - Flags: superreview?(bzbarsky)
Attachment #127689 - Flags: review?(dougt)
Comment on attachment 127689 [details] [diff] [review] v1.1 patch on unsolisted notifications -- there is nothing specifically ruling them out -- but clearly this would be a bigger problem than this bug. I don't see any reason to check for correctness on these pref values.
Attachment #127689 - Flags: review?(dougt) → review+
yeah, strictly checking for valid pref values shouldn't be a requirement; however, it is easy to do in this case without costing us anything.
Comment on attachment 127689 [details] [diff] [review] v1.1 patch >Index: base/src/nsProtocolProxyService.cpp >+ prefsInt->AddObserver("network.proxy", this, PR_FALSE); >Index: base/src/nsProtocolProxyService.h >+ nsCOMPtr<nsIPrefBranch> mPrefs; This gives you a reference cycle. >Index: streamconv/converters/nsDirIndexParser.cpp >+ if (!defCharset.IsEmpty()) >+ mEncoding.Assign(NS_ConvertUCS2toUTF8(defCharset).get()); CopyUTF16toUTF8(defCharset, mEncoding); sr=me with those fixed.
Attachment #127689 - Flags: superreview?(bzbarsky) → superreview+
fixed-on-trunk
Status: NEW → RESOLVED
Closed: 21 years ago
Resolution: --- → FIXED
the checkin here added a printf to nsProtocolProxyService::PrefsChanged (I hit it every time). the patch to nsProtocolProxyService.cpp seems to do a lot of stuff not contained in attachment 127689 [details] [diff] [review]... http://bonsai.mozilla.org/cvsview2.cgi?diff_mode=context&whitespace_mode=show&subdir=mozilla/netwerk/base/src&command=DIFF_FRAMESET&file=nsProtocolProxyService.cpp&rev1=1.47&rev2=1.48&root=/cvsroot
Andrew: the diffs between the current patch and the one that got checked in are based on bz's comments (we discussed the diffs over IRC). sorry about the errant printf!
You need to log in before you can comment on or make changes to this bug.

Attachment

General

Creator:
Created:
Updated:
Size: