On Tue, Nov 19, 2002 at 06:14:58PM +0100, Balazs Scheidler wrote:
First of all you have linked syslog-ng 1.4 against libol 0.3. This is _not_ recommended (and can only be done by patching syslog-ng configure script)
Oh! If I did, it must have happened by accident - but I certainly did not change anything in the configure-script.
The disassembly of your do_flush function shows (at least to me and I was watching it quite a lot) that you might have compiled libol with a buggy compiler. I might be wrong, but the memory slot referenced when calling c_format() is not initialized by the assembly function in any way (it is but in a later part of the function), but it is initialized by the C code.
Don't you happen to be using the buggy gcc 2.96 release from RedHat?
Now that you mention it: [root@goldfinger redhat]# gcc -v Reading specs from /usr/lib/gcc-lib/i386-redhat-linux/2.96/specs gcc version 2.96 20000731 (Red Hat Linux 7.3 2.96-110)
The relevant part is: [...] as you can see
No, I don't ;-) I'm just a regular sysadmin, and can't make head nor tail of assembly language, but thanks anyway - sounds like a brilliant analysis. And sorry to have wasted your time... DeadRat can be a pain in the a**, sometimes. Heinz Ekker