300877 - beachball on button click in page

Status: RESOLVED WORKSFORME

(Camino Graveyard :: General, defect)

Description

[Julian Pellico]

OS X 10.4.0
Camino 2005070308

Camino can enter the "beachball" state (and never leave it) after clicking on a
submit button on a page. It can even happen on a 'Try again' button on a Camino
error page. The beachball happens before the page content changes at all.

I'll post a bunch of samples taken with Thread Viewer while the beach balling
happens.

I forgot to check the CPU/memory usage while it happened. I'll do that if it
happens again.

Comment 1

[Julian Pellico]

Attachment: random stack traces

Comment 2

[Mike Pinkerton (not reading bugmail)]

um. ok? this works for me?

Comment 3

[Simon Fraser [no longer active]]

Can you attach sampler data? Those stack traces don't show much.

Also, what pages do you see this on? What other pages are open in other tabs? Is
there lots of Flash or Java running?

Comment 4

[Julian Pellico]

At the occurrence yesterday, there was only one tab open containing the error page.

Comment 5

[Julian Pellico]

(In reply to comment #2)
> um. ok? this works for me?

Ah, well if it works for Mike, then there's no issue =)

Comment 6

[Julian Pellico]

Attachment: Sampler file (.trace.gz)

It just happened twice somehow - once on the Camino error page agin. Once when
I clicked the Submit button to post this new attachment to this bug :(

I pruned the sample root to the thread that appears to be the drawing thread.

Comment 7

[Julian Pellico]

Oh, sampler saves all the info in the trace - you can set the root yourself if
you want :)

Comment 8

[Simon Fraser [no longer active]]

I dont' see anything unusual in the trace.

Is Camino using lots of CPU when it beackballs, or does the CPU usage drop to zero?
Can you run ThreadViewer when it beachballs, and, if there are active threads
(green) get stacks for them. If there are deadlocked threads (pink) gets stacks
for them too (by just clicking on the thread).

Comment 9

[Julian Pellico]

I watched Thread Viewer the first time it happened. Exactly one of the threads
was running (green). That thread produced the changing stack traces in the first
attachment.

I checked the CPU usage today and Camino took 90%. The system didn't feel
sluggish, though.

Comment 10

[Julian Pellico]

I haven't seen this since I installed a later nightly on 20050717.