Linux: Add Breakpad client.

Linux: Add Breakpad client.

This adds a Breakpad client in breakpad/linux. Breakpad already
contains a Linux client however, for many reasons, it is not of
sufficient quality to use in Chrome. Nor can it easilly be patched
into shape.

This change also forks minidump_file_writer.cc from the Breakpad
source in order to remove it's dependency on the dynamic linker. This
is a small forking an should be upstreamed when we have time.

BUG=10772

[agl, 2009-05-21 01:14:03]

UNFINISHED:

I haven't managed to do my own read-though on this code yet, but I'm out
tomorrow morning/maybe afternoon so I thought I'd throw it out there in case you
wanted to take a look now.

It builds and seems to work. The upload URL and version number are currently
hardcoded to testing values, which I'll need to change before landing.

Hopefully you can find the debugging hacks that I've forgotten about at least.


Cheers

[Evan Martin, 2009-05-21 16:55:20]

I started at the top and only got a few files in before I was preempted.

I think Markus is probably the right person to review this.

It'd be helpful to stick a README in the dir mentioning design points (e.g.
"can't allocate memory").

http://codereview.chromium.org/115526/diff/1025/27
File breakpad/linux/directory_reader.h (right):

http://codereview.chromium.org/115526/diff/1025/27#newcode59
Line 59: // entry over and over.
In light of the above comment it might be clearer to have this called GetEntry
and describe it in terms of the "current" entry, where it returns false when
there is no current.

http://codereview.chromium.org/115526/diff/1025/29
File breakpad/linux/exception_handler.cc (right):

http://codereview.chromium.org/115526/diff/1025/29#newcode61
Line 61: // context and are easy the code, others run in a compromised context
and the
typo: "easy the code"

http://codereview.chromium.org/115526/diff/1025/29#newcode131
Line 131: // Runs before crashing: normal context.
these comments are very helpful, thanks!

http://codereview.chromium.org/115526/diff/1025/29#newcode158
Line 158: // mask all exception signals which we're handling one of them.
grammar typo

http://codereview.chromium.org/115526/diff/1025/30
File breakpad/linux/exception_handler.h (right):

http://codereview.chromium.org/115526/diff/1025/30#newcode189
Line 189: // A vector of the old signal handlers. The void* is a pointer to be
new
"to be new" is a typo somewhere.  I think this is trying to say it's a poitner
to the old sigaction structure?  (Or maybe a newly allocated one with the old
handler's details?)