curl-users
the curl bug (fwd)
From: Daniel Stenberg <daniel_at_haxx.se>
Date: Mon, 25 Sep 2000 09:42:21 +0200 (MET DST)
Date: Mon, 25 Sep 2000 09:42:21 +0200 (MET DST)
Hey ho.
I'm struggling hard with a backlog of curl mails. Here's one of the reports
I've got. Feel free to investigate, comment, suggest. Whatever.
(I removed Vladi's email address below since he mailed me this privately)
--
Daniel Stenberg - http://daniel.haxx.se - +46-705-44 31 77
ech`echo xiun|tr nu oc|sed 'sx\([sx]\)\([xoi]\)xo un\2\1 is xg'`ol
---------- Forwarded message ----------
Date: Thu, 21 Sep 2000 15:20:19 +0300 (EEST)
From: Vladi Shabanski
To: daniel_at_haxx.se
Subject: the curl bug
hi!
You remember I contacted you once before about a rare (strange) bug
in curl. Now I think I found the exact situation which triggers it.
Unfortunately I couldnt make it to debug it myself -- gdb says it
cannot set/find _Transfer() breakpoint and some weird messages :)
I removed -O2 but it didn't help... however this is it:
I take 3 text files:
-rw-r--r-- 1 cade users 100000 Sep 21 14:36 l01.txt
-rw-r--r-- 1 cade users 5000 Sep 21 14:36 l02.txt
-rw-r--r-- 1 cade users 2000 Sep 21 14:37 l03.txt
I put them in my public_html on the local machine:
/home/cade/public_html/l*.txt
Then start curl:
curl -O http://localhost/~cade/l[01-03].txt
First file is downloaded ok, second starts (i.e. curl just reports
0% downloaded stats) and it segfaults right there! AFAI can find the
situation is that _Transfer( conn* ) gets NULL and in speedcheck( data*)
somewhere around line 500 or something data* gets random value
sometimes it is NULL other some invalid address (0x1 etc...)
It never gets to the third file...
I can reproduce this error clearly with latest stable version:
curl 7.2 (i586-pc-linux-gnu) libcurl 7.2.1
and with 6.5.2
I think it is important that second file is smaller that first one
and that first one is about 100K, I mean that with files with sizes
of 4K and 1K (for example) the bug don't happen! I couldn't reproduce
it neither with equally sized files nor in reverse order i.e. smaller
first.
I hope this will finally (and hopefully) help to fix this bug!
If you still cannot reproduce it (which I really don't believe), I
can send you my curl sources l*.txt files and anything you think
can help...
Thanx for the attention and good luck!
P! Vladi.
--
Vladi Belperchinov-Shabanski
http://www.biscom.net/~cade
DataMax Ltd.
Received on 2000-09-25