Re: pkg/32800: wine can't create ~/.wine

[dholland%eecs.harvard.edu@localhost (David Holland)]

The following reply was made to PR pkg/32800; it has been noted by GNATS.

From: dholland%eecs.harvard.edu@localhost (David Holland)
To: gnats-bugs%NetBSD.org@localhost
Cc: 
Subject: Re: pkg/32800: wine can't create ~/.wine
Date: Sun, 22 Oct 2006 18:17:11 -0400 (EDT)

 I've found another problem... unfortunately it's still not the whole
 problem.
 
 NetBSD's libpthread ignores the per-thread stack size that can be set
 with pthread_attr_setstacksize(), and it looks like Wine relies on
 this in some fashion. It's not clear to me exactly what's going on,
 but:
    - when it dumps core, the current thread's stack pointer is 0x7fcefe8c;
    - pthread__self() constructs 0x7c000000 from this because
      pthread_stacksize is 0x4000000;
    - however, the bottom of the current stack as it exists in memory,
      as ascertained by poking memory with gdb, is 0x7fbe0000;
    - wine uses pthread_attr_setstacksize() - see wine-0.9.23/loader/pthread.c.
 
 I tried working around this by setting $PTHREAD_STACKSIZE to 1152
 (0x120000 / 1024) in the hopes that this would cause pthread__self to
 return the "right" address. This allows wineprefixcreate, or rather,
 the wine invocation inside it, to get past the initial SIGSEGV, but it
 still doesn't work - it appears to be back to the infinite loop I was
 seeing last spring after the signal fixes.
 
 Sigh.
 
 -- 
    - David A. Holland / dholland%eecs.harvard.edu@localhost