Move DumpBacktrace() to checks.cc and cleanup both the code and the necessary platform checks.

Move DumpBacktrace() to checks.cc and cleanup both the code and the necessary platform checks.

This also removes the platform-posix.h header file.

R=machenbach@chromium.org

Committed: https://code.google.com/p/v8/source/detail?r=16890

[Benedikt Meurer, 2013-09-23 09:54:45]

Hey Michael,
This moves the DumpBacktrace() function to checks.cc, and cleans up the platform
checks (actually *BSD and glibc only, which was difficult to figure out before).
PTAL
-- Benedikt

[Michael Achenbach, 2013-09-23 11:12:50]

https://codereview.chromium.org/23497009/diff/1/src/checks.cc
File src/checks.cc (right):

https://codereview.chromium.org/23497009/diff/1/src/checks.cc#newcode30
src/checks.cc:30: #if V8_LIBC_GLIBC || V8_OS_BSD
Does that include !defined(__UCLIBC__)?

https://codereview.chromium.org/23497009/diff/1/src/checks.cc#newcode84
src/checks.cc:84: fflush(stderr);
Why can that not stay within DumpBacktrace()?

https://codereview.chromium.org/23497009/diff/1/src/platform-macos.cc
File src/platform-macos.cc (left):

https://codereview.chromium.org/23497009/diff/1/src/platform-macos.cc#oldcode68
src/platform-macos.cc:68: // This lets us launch on 10.4 which does not have
these calls.
In the new code, execinfo.h will be included. Will it still work on 10.4?

https://codereview.chromium.org/23497009/diff/1/src/platform-posix.h
File src/platform-posix.h (left):

https://codereview.chromium.org/23497009/diff/1/src/platform-posix.h#oldcode60
src/platform-posix.h:60: #if !defined(ANDROID)
What if we are on Android in the new code? Why is it safe to remove the #if?

[Benedikt Meurer, 2013-09-23 11:26:40]

https://codereview.chromium.org/23497009/diff/1/src/checks.cc
File src/checks.cc (right):

https://codereview.chromium.org/23497009/diff/1/src/checks.cc#newcode30
src/checks.cc:30: #if V8_LIBC_GLIBC || V8_OS_BSD
On 2013/09/23 11:12:50, machenbach wrote:
> Does that include !defined(__UCLIBC__)?

Jep, our V8_LIBC_* are strict. I.e. V8_LIBC_GLIBC is only set for glibc.

https://codereview.chromium.org/23497009/diff/1/src/checks.cc#newcode84
src/checks.cc:84: fflush(stderr);
On 2013/09/23 11:12:50, machenbach wrote:
> Why can that not stay within DumpBacktrace()?

Because we should flush stderr on either case.

https://codereview.chromium.org/23497009/diff/1/src/platform-macos.cc
File src/platform-macos.cc (left):

https://codereview.chromium.org/23497009/diff/1/src/platform-macos.cc#oldcode68
src/platform-macos.cc:68: // This lets us launch on 10.4 which does not have
these calls.
On 2013/09/23 11:12:50, machenbach wrote:
> In the new code, execinfo.h will be included. Will it still work on 10.4?

Nope. We've moved to 10.6 as announced publicly last week.

https://codereview.chromium.org/23497009/diff/1/src/platform-posix.h
File src/platform-posix.h (left):

https://codereview.chromium.org/23497009/diff/1/src/platform-posix.h#oldcode60
src/platform-posix.h:60: #if !defined(ANDROID)
On 2013/09/23 11:12:50, machenbach wrote:
> What if we are on Android in the new code? Why is it safe to remove the #if?

The code requires glibc or BSD. Neither of which includes Android.

[Michael Achenbach, 2013-09-23 12:13:27]

lgtm then...

[Benedikt Meurer, 2013-09-23 14:12:13]

Committed patchset #1 manually as r16890 (presubmit successful).