[Valgrind-developers] Strange behavior of clients started under drd on AMD64 -- bug in VEX ?

[Bart V. A. <bar...@gm...>]

Hello,

    While porting drd from X86 to AMD64 I noticed that all
multithreaded clients I ran under drd were terminated by a SIGSEGV
reported in std::locale::locale(). After some time I found a
workaround: insert the following statement in the pthread_mutex_lock()
wrapper (which does nothing except stopping statement reordering in
VEX):
   write(1, "", 0);
When I saw this crash the first time I expected it to be a bug in drd.
But after I found the workaround I'm no longer sure it's a bug in drd
-- maybe it's a bug in VEX ? Details of how to reproduce the above
behavior can be found in http://bugs.kde.org/show_bug.cgi?id=142070.

-- 
Regards,

Bart Van Assche.