2008/9/3, Tor Arntsen wrote:

> With that patch the http tests started working. It still fails on the
> ftp tests, i.e. it starts failing at test 100:
> [...]
> As seen before, it's the calloc() call itself that appears to be sigsegv'ing.

What a nightmare!!

Assuming that nothing changed on the system on July 9 that could
affect how ICC optimizes our code, I fear that I have no better answer
than suggesting that you checkout CVS at different dates/versions
until you find the most recent one that works.

Something like..

cvs -d :pserver:anonymous_at_cool.haxx.se:/cvsroot/curl checkout -D
"2008-07-09 18:37 UTC" curl

And even after doing that, it might show a clear answer or not.

Could you disable all compiler inlining optimizations while doing these tests ?

An easy question; are these failures happening on a single machine or
on more than one. If its only on one, it could even be a hardware
thing.

-- 
-=[Yang]=-