499446 - Bug 497680 cause reported memory leaks in xpcshell-tests (bayesian-spam-filter and test_bug366491.js)

Status: RESOLVED FIXED

(MailNews Core :: Backend, defect)

Description

[Mark Banner (:standard8)]

The landing of bug 497680 caused reported leaks in all the extensions/bayesian-spam-filter tests and test_bug366491.js (mailnews/base/test/unit).

This is possibly similar to our other rdf resource leak bugs, e.g. bug 498323 or bug 498321), however reporting it here as it is a fresh reported leak.

Comment 1

[David :Bienvenu]

I'll look at this.

Comment 2

[David :Bienvenu]

it looks like a few of the local folder special folders are leaking, e.g., Trash, Unsent Messages...probably a ref-counting problem...

Comment 3

[David :Bienvenu]

Attachment: proposed fix

fixes ref counting problem in GetAllFolders.

The closer we get to 0 leaks, the easier it is to find leak regressions :-)

Comment 4

[Mark Banner (:standard8)]

Comment on attachment 384265 [details] [diff] [review]
proposed fix

>   for (i = 0; i < folderCount; i++)
>-    folderArray->AppendElement(allDescendents->ElementAt(i), PR_FALSE);
>+  {
>+    nsCOMPtr<nsIMsgFolder> folder(do_QueryElementAt(allDescendents, i));
>+    folderArray->AppendElement(folder, PR_FALSE);
>+  }

I think moving the nsCOMPtr outside the loop is meant to be a bit more efficient. Alternately we could just drop the nsCOMPtr and append the result of do_QueryElementAt, as the change to do_QueryElementAt is what is fixing the leak here.

r/sr=Standard8 either way.

Comment 5

[David :Bienvenu]

do_QueryElementAt by itself doesn't compile - perhaps some sort of casting would have helped, but I just moved the comptr outside the loop. Fix checked in.