34310 - assert: !"I/O method is invalid" at priometh.c:65

Status: VERIFIED WORKSFORME

(Core :: Networking, defect, P3)

Description

[Kathleen :Brade]

Running a build from this morning, I get the following assert:
!"I/O method is invalid" at priometh.c:65
The stack crawl (partial) looks like this:
  PR_Read
  nsSocketTransport::IsAlive(unsigned in, int*)+00090
  nsHTTPHandler::RequestTransport(nsIRUI* nsHTTPChannel*, unsigned int, unsigned
int, nsIChannel**, unsigned int)+00440
  nsHTTPPipelinedRequest::WriteRequest()+00154
  nsHTTPChannel::Open()+00B24
  nsHTTPHandler::ReleaseTransport(nsIChannel*, unsigned int)+003FC
  nsHTTPServerListener::OnStopRequest(nsIChannel*, nsISupports*, unsigned int,
const unsigned short*)+003A8
  ...

Comment 1

[ruslan]

Is it happening all the time? The stack looked as it normally should.

Comment 2

[ruslan]

warren, gordon - you've MACs. Can you take a look real quick? IsAlive method was 
there for a while and I didn't touch it recently. Can you set a breakpoint in 
IsAlive (in PR_Read) and see if fd looks ok?

Comment 3

[Kathleen :Brade]

I wrote up this bug report when it first happened so I'd be able to provide the
stack information.  Unfortunately I haven't been able to reproduce this on a
second attempt.

Comment 4

[ruslan]

I just finished my brand-new build and I can't even load www.cnn.com - there's 
something very wrong with layout and pages keeps on reloading all the time (on 
NT). It wasn't the case on Friday. I think we're looking into some unrelated 
bustage somewhere.

Comment 5

[Warren Harris]

=> gordon

Comment 6

[Warren Harris]

Moving to M17 which is now considered part of beta2.

Comment 7

[leger]

Can PDT get a retest on this please?  Need to know results with latest build.  
Thanks!

Comment 8

[gordon]

Kathy, have you seen this since it was reported.  I've never seen this assert in 
any of my builds.

Comment 9

[Kathleen :Brade]

no, I haven't seen this bug since I filed it

Comment 10

[gordon]

Okay, I'm going to mark this one WORKSFORME.  If we see it again, we'll take care 
of it then.

Comment 11

[leger]

Marking Verified.