Re: Someone stomping on our memory

Andres Kroonmaa wrote:
>On 20 Nov 2000, at 8:50, Stephen R. van den Berg <srb@cuci.nl> wrote:
>> I hate to say it, but I squashed these leaks in my code more than two
>> years ago (which proves that you didn't look at my patch 8-).

> hmm. weird. to my knowledge 2 years ago we didn't have ftp.c on pools.
> I came across this bug when I started to look close into 2.4. 2.3 did
> not have this bug, as it didn't have mempools in ftp.c

Well, 2.2 did have it, apparently, otherwise my patch would have shown more
differences.

> I did look at your patch, but only in relation to chunked pools. I hope
> you agree that it is quite unexpected to find a bug fix to bug in a patch
> to s2.2, and to bug introduced between 2.3 and 2.4...

Mempools existed earlier than 2.3.

>> From my old diffs:

>> diff -U 2 -b -B -p -r -d --horizon-lines=2 -X /usr/local/etc/xdiff squid-2.2.STA
>> BLE3/src/ftp.c squid-B2/src/ftp.c
>> --- squid-2.2.STABLE3/src/ftp.c Tue May 11 22:34:36 1999
>> +++ squid-B2/src/ftp.c Sun Jun 27 04:52:54 1999

>> So it appears that you might have missed another spot still.
>> I didn't compare it with the current code yet.

> there are many changes in your patch that are most probably not compatible
> with current state, therefore it is very difficult to judge whether any
> part of it may fix our current problems.

True, perhaps. Though the mempool leaks kind of sticked out.

>@@ -1179,5 +1217,5 @@ clientCacheHit(void *data, char *buf, ss
> debug(33, 3) ("clientCacheHit: request aborted\n");
> return;
>- } else if (size < 0) {
>+ } else if (size <= 0) {
> /* swap in failure */
> memFree(buf, MEM_CLIENT_SOCK_BUF);

> I can see few places in your patch that may be related to crashes I'm
> seeing, but I'm unable to see how they interact with other areas you
> address, so its difficult to say whether they are applicable.

Some of those, like the one you singled out above, are standalone patches
which should be generic fixes.

> Perhaps you would care to take 2.4 and review you changes, and extract
> those that could fix any bugs we've missed?

I will, as soon as I can spare some time.

-- 
Sincerely,                                                          srb@cuci.nl
           Stephen R. van den Berg (AKA BuGless).
I'm sorry. The number you have reached is imaginary.
Please rotate your phone 90 degrees and try again.