Add frame pointer recovery to the AMD64 Stackwalker.

Add frame pointer recovery to the AMD64 Stackwalker.

BUG=https://code.google.com/p/chromium/issues/detail?id=393594
R=mark@chromium.org

Committed: https://code.google.com/p/google-breakpad/source/detail?r=1350

[Mark Mentovai]

LGTM

https://breakpad.appspot.com/10664002/diff/1/src/processor/stackwalker_amd64.cc
File src/processor/stackwalker_amd64.cc (right):

https://breakpad.appspot.com/10664002/diff/1/src/processor/stackwalker_amd64....
src/processor/stackwalker_amd64.cc:156: // Assume the presense of a frame
pointer.
presence

https://breakpad.appspot.com/10664002/diff/1/src/processor/stackwalker_amd64....
src/processor/stackwalker_amd64.cc:156: // Assume the presense of a frame
pointer.
Add something to this comment that says that the x86_64 ABI does not mandate the
use of the %rbp frame pointer, but in the absence of CFI, attempting FP recovery
is better than stack scanning.

The x86_64 ABI document is http://www.x86-64.org/documentation/abi.pdf. I’m
talking about section 3.2.2, especially footnote 7.

[rsesek]

https://breakpad.appspot.com/10664002/diff/1/src/processor/stackwalker_amd64.cc
File src/processor/stackwalker_amd64.cc (right):

https://breakpad.appspot.com/10664002/diff/1/src/processor/stackwalker_amd64....
src/processor/stackwalker_amd64.cc:156: // Assume the presense of a frame
pointer.
On 2014/07/18 00:14:32, Mark Mentovai wrote:
> Add something to this comment that says that the x86_64 ABI does not mandate
the
> use of the %rbp frame pointer, but in the absence of CFI, attempting FP
recovery
> is better than stack scanning.
> 
> The x86_64 ABI document is http://www.x86-64.org/documentation/abi.pdf. I’m
> talking about section 3.2.2, especially footnote 7.

Done.

https://breakpad.appspot.com/10664002/diff/1/src/processor/stackwalker_amd64....
src/processor/stackwalker_amd64.cc:156: // Assume the presense of a frame
pointer.
On 2014/07/18 00:14:32, Mark Mentovai wrote:
> presence

Done.

https://breakpad.appspot.com/10664002/diff/1/src/processor/stackwalker_amd64....
src/processor/stackwalker_amd64.cc:156: // Assume the presense of a frame
pointer.
On 2014/07/18 00:14:32, Mark Mentovai wrote:
> presence

Done.

[Mark Mentovai]

LGTM

https://breakpad.appspot.com/10664002/diff/10004/src/processor/stackwalker_am...
File src/processor/stackwalker_amd64.cc (right):

https://breakpad.appspot.com/10664002/diff/10004/src/processor/stackwalker_am...
src/processor/stackwalker_amd64.cc:158: // compilers to still emit this
information.
Nit: the compilers aren’t emitting information, they’re using %rbp in this way
and emitting instructions that preserve %rbp in a consistent location on the
stack.

[rsesek]

https://breakpad.appspot.com/10664002/diff/10004/src/processor/stackwalker_am...
File src/processor/stackwalker_amd64.cc (right):

https://breakpad.appspot.com/10664002/diff/10004/src/processor/stackwalker_am...
src/processor/stackwalker_amd64.cc:158: // compilers to still emit this
information.
On 2014/07/18 00:25:25, Mark Mentovai wrote:
> Nit: the compilers aren’t emitting information, they’re using %rbp in this way
> and emitting instructions that preserve %rbp in a consistent location on the
> stack.

Done.

[rsesek]

Committed patchset #3 manually as r1350 (presubmit successful).