Jan Kratochvil wrote:
Hi,
On Fri, 04 Apr 2008 04:53:58 +0200, Roland McGrath wrote:
I'd like to make sure this all seems coherent and there aren't
holes in my logic you can think of before I send this upstream.
It all PASSes all now on x86_64 and i686 (except step-to-breakpoint) on (files
will vanish in several days as it is in /scratch/):
http://koji.fedoraproject.org/scratch/jkratoch/task_547885/
step-jump-cont-strict PASS on utrace patched 2.6.25 x86 kernel, fails on
on utrace patched 2.6.25 x86_64 kernel. It says:
./step-jump-cont-strict: step-jump-cont-strict.c:501: test_singleblock
(, instruction -1): IP 0x400917 expected: found 0x400911
I found the definition of PTRACE_SINGLEBLOCK in test case affects the
result.
If we remove the PTRACE_SINGLEBLOCK definition statements from
__x86_64__ branch, the test will PASS.
Vice versa, adding the PTRACE_SINGLEBLOCK definition to __i386__ branch,
the test will fail.
./step-jump-cont-strict: step-jump-cont-strict.c:504: test_single block
(, instruction -1): IP 0x804870f expected: found 0x8048709
In fact, the x86/x86_64 kernel share the same PTRACE_SINGLEBLOCK in
asm-x86/ptrace-abi.h. Seems that we have
the different expectations about x86 and x86_64?
Regards,
Wenji
