52910 - Crash involving JavaScript alert in one thread and incoming page in another

Status: VERIFIED WORKSFORME

(Core :: Networking, defect, P3)

Description

[Bob Kanefsky]

Build id:  2000090811
Reproducible:  Haven't tried.  Suspect it's sensitive to timing.

I was testing the following JavaScript code:

function goto_page (url) {
   location.replace(url);
   setTimeout("retry_goto_page('" + url + "')", server_response_timeout);
}

function retry_goto_page (url) {
   if (confirm ("Sorry for the delay.  Our server is probably either very busy 
or is down.  Try again?"))
     { goto_page(url) }
}

I caused the alert to come up by deliberately delaying my server's response.  
Then I allowed the server to respond, and the page started loading 
asynchronously behind the dialog.  I clicked on the OK button (I think), causing 
it to location.replace with the same URL that was in the middle of loading.  
I'll attach a MacsBug StdLog.  Mozilla was the selected application and shows up 
in the stack trace, but note that the log shows Macintosh Common Lisp as the 
currently running process.

Comment 1

[Bob Kanefsky]

Attachment: MacsBug StdLog

Comment 2

[Frank Tang]

Bob Kanefsky- your test cases is very difficult to reproduce. Is that possible
you can put it into a html page so other developers can reproduce it easiler ?

Comment 3

[Bob Kanefsky]

Very well.  I've put HTML around it, and also put it on an HTTP server.

http://shuttle-upgrade.arc.nasa.gov/mozilla-bug/52910/page1

I never said it would be easy to reproduce.

Comment 4

[Frank Tang]

I try both win/mac 10/16/2000 branch build and I cannot reproduce crash.
bob- can you still crash it ?

Comment 5

[Bob Kanefsky]

No, it only happened once.  Thread errors are like that.  We had one on the 
Remote Agent Experiment last year that never showed up in our hundreds of hours 
of testing, and only happened when it was in space.  Someone forgot to wrap 
"(with-critical-section ...)" around some code, and it decided the ion drive 
wasn't ready to start thrusting.

Comment 6

[Darin Fisher]

added keyword qawanted.

Comment 7

[Fabian Guisset]

This would be a Dup of bug 56337 which has been fixed on 2000/11/29.
Please verify.
Fabian.

Comment 8

[Bob Kanefsky]

I can't reproduce the timing problem that set it off, even under an old build 
(2000080712) -- but then, I can't reproduce bug 56337 with that build either on 
MacOS 9.0.4.  The stack trace doesn't resemble anything else, except another Mac 
bug (bug 29908) that was fixed months before this bug cropped up.

So there's no evidence either way as to whether it's a duplicate or is fixed.  
The patch that fixes bug 56337 doesn't directly affect the code on my stack 
trace, so only someone familiar with the thread library could say whether it's 
plausible that it prevents the error this bug report captured.  I can't determine 
the question one way or the other.

Comment 9

[Darin Fisher]

marking WORKSFORME... please reopen if problem re-surfaces.

Comment 10

[Tom Everingham]

worksforme too, verified Mac os8.6 2001051504