Fix cctest/test-cpu-profiler/CollectCpuProfile test on Arm and MIPS simulators

Fix cctest/test-cpu-profiler/CollectCpuProfile test on Arm and MIPS simulators

Signal handler on simulator now retrieve registers from the simulator not from the host machine.

BUG=v8:2621
Committed: https://src.chromium.org/viewvc/chrome?view=rev&revision=14235

[Sven Panne, 2013-04-11 09:58:02]

Just a few drive by comments. I think we should consistently add the changes in
platform-linux.cc to platform-{solaris,freebsd,openbsd}.cc, too. It's just a
matter of copy-n-paste and it's nice to keep them similar (I should really start
to rearrange the platform support... :-P ).

https://codereview.chromium.org/13845014/diff/1/src/platform-linux.cc
File src/platform-linux.cc (right):

https://codereview.chromium.org/13845014/diff/1/src/platform-linux.cc#newcode...
src/platform-linux.cc:1102: #if V8_TARGET_ARCH_ARM || V8_TARGET_ARCH_MIPS
I think we can remove this #if...

https://codereview.chromium.org/13845014/diff/1/src/platform-linux.cc#newcode...
src/platform-linux.cc:1124: sample->fp =
reinterpret_cast<Address>(sim->get_register(Simulator::fp));
... and add an #else followed by #error here.

https://codereview.chromium.org/13845014/diff/1/test/cctest/cctest.status
File test/cctest/cctest.status (left):

https://codereview.chromium.org/13845014/diff/1/test/cctest/cctest.status#old...
test/cctest/cctest.status:80: # BUG(2621): Stack traversal doesn't work properly
on Arm simulator.
I've already disabled it for MIPS, too, so this should be removed then after
rebasing this CL, too.

[Jakob Kummerow, 2013-04-11 10:01:33]

LGTM when Sven's comments are addressed.

[yurys, 2013-04-11 13:00:10]

Added support for simulator on other platforms. As discussed off the issue I
left #if V8_TARGET_ARCH_ARM || V8_TARGET_ARCH_MIPS untouched as there are some
rudimentary simulator support for ia32/x64 which would be broken by the
suggested changes. PTAL.

On 2013/04/11 09:58:02, Sven Panne wrote:
> Just a few drive by comments. I think we should consistently add the changes
in
> platform-linux.cc to platform-{solaris,freebsd,openbsd}.cc, too. It's just a
> matter of copy-n-paste and it's nice to keep them similar (I should really
start
> to rearrange the platform support... :-P ).
> 
> https://codereview.chromium.org/13845014/diff/1/src/platform-linux.cc
> File src/platform-linux.cc (right):
> 
>
https://codereview.chromium.org/13845014/diff/1/src/platform-linux.cc#newcode...
> src/platform-linux.cc:1102: #if V8_TARGET_ARCH_ARM || V8_TARGET_ARCH_MIPS
> I think we can remove this #if...
> 
>
https://codereview.chromium.org/13845014/diff/1/src/platform-linux.cc#newcode...
> src/platform-linux.cc:1124: sample->fp =
> reinterpret_cast<Address>(sim->get_register(Simulator::fp));
> ... and add an #else followed by #error here.
> 
> https://codereview.chromium.org/13845014/diff/1/test/cctest/cctest.status
> File test/cctest/cctest.status (left):
> 
>
https://codereview.chromium.org/13845014/diff/1/test/cctest/cctest.status#old...
> test/cctest/cctest.status:80: # BUG(2621): Stack traversal doesn't work
properly
> on Arm simulator.
> I've already disabled it for MIPS, too, so this should be removed then after
> rebasing this CL, too.

[Jakob Kummerow, 2013-04-11 13:07:07]

lgtm

[Sven Panne, 2013-04-11 13:11:18]

Still one nit...

https://codereview.chromium.org/13845014/diff/10001/src/platform-freebsd.cc
File src/platform-freebsd.cc (right):

https://codereview.chromium.org/13845014/diff/10001/src/platform-freebsd.cc#n...
src/platform-freebsd.cc:721: Isolate::PerIsolateThreadData* per_thread_data =
Can we stash the thread id into PlatformData on *nix-like platforms too? It
would be great for consistency and easier cleanup/refactoring of the platform
support. Currently it is very hard to find out if a difference is fundamental or
just an oversight.

[Sven Panne, 2013-04-11 13:25:30]

LGTM when the PlatformData changes are done (as discussed offline)

[yurys, 2013-04-11 14:16:05]

https://codereview.chromium.org/13845014/diff/10001/src/platform-freebsd.cc
File src/platform-freebsd.cc (right):

https://codereview.chromium.org/13845014/diff/10001/src/platform-freebsd.cc#n...
src/platform-freebsd.cc:721: Isolate::PerIsolateThreadData* per_thread_data =
On 2013/04/11 13:11:18, Sven Panne wrote:
> Can we stash the thread id into PlatformData on *nix-like platforms too? It
> would be great for consistency and easier cleanup/refactoring of the platform
> support. Currently it is very hard to find out if a difference is fundamental
or
> just an oversight.

Done. Changed implementation on all platforms to use ThreadId saved in
PlatformData.

[yurys, 2013-04-11 14:22:25]

Committed patchset #6 manually as r14235.