Google Code Exporter

``` Meta: https://bugzilla.mozilla.org/show_bug.cgi?id=19119 Since this is timing related, we should fix issue 162 first, and see what difference that makes. ``` Original comment by `[email protected]` on 24 Apr 2011 at...

``` In the meantime, altered the DNS to point directly to floodgap.com and let HTTPi sort it out. This wallpapers the problem at least for us. tenfourfox.com will now become...

``` Theory, aided in part by http://www.mactech.com/articles/develop/issue_27/macqa.html This is a fairly subtle race condition in NSPR. Looking at nsprpub/pr/src/md/mac/macsockotpt.c ; our OpenTransport notifier is in NotifierRoutine(). 1. We connect, transmit...

``` I was hoping Copperweb could let me simulate this on my internal network but no dice. ``` Original comment by `[email protected]` on 1 Jul 2011 at 4:34

``` Well, did some initial work on this. First, there is already an fd->secret->orderlyDisconnect and we will simply (ab)use that as our signal to close the socket within SendReceiveStream(). So...

Original comment by `[email protected]` on 1 Jul 2011 at 5:20 - Added labels: **Milestone-Release9.2.3**

``` It's not T_ORDREL. It's T_DISCONNECT, with a disconnect error of 54. This is ECONNRESET = 54; { Connection reset by peer } (which makes sense). We know IE can...

``` So the current approach hasn't worked at all; we still have a serious race condition. Realistically all I can think of is to build a queue of network states...

``` For the queue, it should consist of the event type (T_ORDREL, T_DISCONNECT, etc.), any useful info, a bookmark and a pointer to a buffer. The queue should be circular...

``` er, copies from the bookmark to the end of the buffer, but OTFreeMems the whole buffer ``` Original comment by `[email protected]` on 30 Aug 2011 at 3:51