126444 - Mozilla dies when jsperf eats up my free disk 32MB 133MHz PC

Status: VERIFIED DUPLICATE of bug 13350

(Core :: DOM: Core & HTML, defect)

Description

[[e-mail]]

Using Build ID: 2002021803 on a 32MB 133MHz PC
I tried running the jsperf test from this page and after Mozilla ate up my free
disk space (64MB) with swapfile it just died, and since I had no memory left I
can't even give you any debugging information.

Comment 1

[Phil Schwartau]

For reference purposes, I will attach the source of the actual performance 
engine, "perf1.html". If you load this page directly, the test will start
to run! Use the view-source protocol to avoid that:

          view-source:http://www.umsu.de/jsperf/perf1.html

Comment 2

[Phil Schwartau]

Attachment: perf1.html (as text)

Comment 3

[Phil Schwartau]

I tried the performance test with Mozilla trunk binary 2002021413 WinNT.
I had no trouble whatsoever. I have a 500 MHz processor and 128M of RAM.
The WinNT Task Manager showed my "available physical memory" dropping from
43M to about 20M at the low point, just before the colorful DHMTL elements
were painted. During most of the run, however, the figure was over 30M.

Note: as the site says, half of the test is JS performance, half is
DHTML performance. Neil: did your browser ever display the colored items
and the little box that travels from the left one to the right? 
Or didn't it get that far?

Comment 4

[[e-mail]]

It didn't get as far as the DHTML performance test. Another 128MB PC did.

Comment 5

[Phil Schwartau]

Neil:

1. On the 32MB box, what is the last message you get in the status bar
   before it dies? That will tell us what the last successful test was -

2. On the 128MB box: does it complete all the tests OK, like mine does?

Comment 6

[[e-mail]]

The 128MB box works fine.

Trying the 32MB box again crashed with the following display:
Title: JavaScript Performance Test - Mozilla {Build ID: 2002021903}
Address: http://www.umsu.de/jspref/perf2.html?1014139515
Testing Core Performance
Please be patient...
This may take about 1 minute...
Mozilla
This program has performed an illegal operation and will be shut down.
Close
If the problem persists, contact the program vendor.
Details
MOZILLA caused an invalid page fault in module KERNEL32.DLL at 0177:bff8433e
Registers:
EAX=00000000 CS=0177 EIP=bff8433e EFLGS=00000246
EBX=007ec1b0 SS=017f ESP=00defeec EBP=00defef0
ECX=c159bff0 DS=017f ESI=00000000 FS=0fc7
EDX=bffc9490 ES=017f EDI=04fb6a50 GS=0000
Bytes at CS:EIP:
5e 5d c2 04 00 55 8b ec 81 ec 0c 03 00 00 53 56
Stack dump:
017ed97c 00deff2c 60e7350e 017ed998 04fb6a50
61032c5d 046cfa50 6103268b 04fb6a50 046cfa50
6103263b 6103264a 046cfa50 007ec240 046cfa50
046cfa50
Sending request to www.umsu.de

Comment 7

[[e-mail]]

Attachment: Managed to get a screenshot this time!

Mozilla sat with the same display for several minutes before the crash dialog.

Comment 8

[Brendan Eich [:brendan]]

Why is this bug not a dup of bug 13350?

/be

Comment 9

[Phil Schwartau]

I see what you mean: if the DOM did a better job of scheduling garbage 
collection, then all the memory of the 32MB box would not get used up.

Will mark as duplicate of bug 13350. Reassigning to DOM Level 0 for parity -

Comment 10

[Phil Schwartau]

Marking as duplicate. Neil, you will be cc'ed on the other bug; you'll  
be able to follow its progress and check that it fixes your testcase -

*** This bug has been marked as a duplicate of 13350 ***

Comment 11

[[e-mail]]

Sure thing.