In the last episode (Mar 04), Xavier Beaudouin said:

Le 3 mars 04, à 18:00, Henrik Grubbström (Lysator) @ Pike (-) developers forum a écrit :

...

...

http://bento.freebsd.org/errorlogs/alpha-5-latest/pike72-7.2.492_2.log

Looks like some problem with varargs.

Ok. Now how do I fix that ?

The va_end call is incorrect. Apply this patch (fixed in 7.3 in 2002 but never backported):

Index: module_support.c =================================================================== RCS file: /cvs/Pike/7.2/src/module_support.c,v retrieving revision 1.40 diff -p -u -r1.40 module_support.c --- module_support.c 20 Feb 2001 13:29:51 -0000 1.40 +++ module_support.c 4 Mar 2004 16:31:39 -0000 @@ -314,7 +314,7 @@ PMOD_EXPORT int get_args(struct svalue * va_start(ptr, fmt); ret=va_get_args(s, num_args, fmt, ptr); #ifndef __TenDRA__ - va_end(fmt); + va_end(ptr); #endif /* !__TenDRA */ return ret; }

...

...

http://bento.freebsd.org/errorlogs/alpha-5-latest/pike74-7.4.43.log

This looks like a bug in the thread library.

Same question.

Since that it works on other platforms (sparc64, amd64, ia64 and i386) ?

Not sure about this one. 5-current is a bit unusual wrt threads right now, since amd64, ia64, and i386 use kernel threads but sparc64 and alpha are still on libc_r. What's weird is the "errno=0" at the end of the error message, which should be impossible. If read() fails, errno had better be set to something.