|
From: Julian S. <js...@ac...> - 2006-08-26 08:46:39
|
> > good and if Julian is also happy I don't see why it can't be added now --
> > it's almost all code that shouldn't affect existing tools.
Mostly it looks ok; however I'd prefer to first stabilise drd itself
and to get some more experience and understanding of what it can/can't
do. Anyway, some questions:
------
+/* pthread_t as handled internally in valgrind. */
+typedef UInt PosixThreadId;
Uh, is that a portable assumption, that a Posix thread id is a
32-bit int? Maybe better would be UWord.
------
asm volatile (
- "movl %1, %0\n" /* set tst->status = VgTs_Empty */
+ // "movl %1, %0\n" /* set tst->status = VgTs_Empty */
"movl %2, %%eax\n" /* set %eax = __NR_exit */
"movl %3, %%ebx\n" /* set %ebx = tst->os_state.exitcode */
"int $0x80\n" /* exit(tst->os_state.exitcode) */
etc. So, I'm really not enthusiastic about messing with the
scheduler's state machine. You made some comment about why this
is necessary, but now I can't find that email. Something to do
with clone?
J
|