Extract breakpad symbol files for Linux official builds.

Extract breakpad symbol files for Linux official builds.

These are archived and submitted to the crash server by the official builder
(coming in a separate buildbot CL).

[TVL, 2009-06-12 12:35:09]

+mark for the redirect in an action.

http://codereview.chromium.org/126015/diff/4/1002
File chrome/chrome.gyp (right):

http://codereview.chromium.org/126015/diff/4/1002#newcode2326
Line 2326: '<(PRODUCT_DIR)/chrome', '>', '<@(_outputs)'],
I'm not sure if shell redirection is safe in an action list?  it might depend on
the generator being used.  for mac, i made a script in src/build/mac that
wrapped this up.  (we should get an arg added to dump_sym to set the output
file)

[Michael Moss, 2009-06-12 17:10:41]

Moved symbol dumping to helper script.

[Michael Moss, 2009-06-12 17:11:22]

And added tool (from thestig) to regenerate breakpad signature.

[TVL, 2009-06-12 17:15:02]

lg

[Mark Mentovai, 2009-06-12 17:30:10]

http://codereview.chromium.org/126015/diff/5/1006
File build/linux/dump_app_syms (right):

http://codereview.chromium.org/126015/diff/5/1006#newcode1
Line 1: #!/bin/sh
svn:executable and svn:eol-style?

http://codereview.chromium.org/126015/diff/5/1006#newcode29
Line 29: echo "ERROR: Could not create temp stripped '$INFILE'"
echo blah >& 2 (also for usage()?)

http://codereview.chromium.org/126015/diff/5/1006#newcode34
Line 34: "$DUMPSYMS" "$INFILE" > "$OUTFILE"
In the Mac version of this script, we have an -nt check to avoid wasting time
doing any work when outfile already exists and is newer than infile.

http://codereview.chromium.org/126015/diff/5/1006#newcode43
Line 43: sed -i "1s/[0-9A-F]* $INFILE_BASE/$NEWSIG $INFILE_BASE/" "$OUTFILE"
This is slightly dangerous in the general case because INFILE_BASE has not
really been scrubbed for sanity.  This should be fine if the sed script is
simply "1s/^[0-9A-F]* /$NEWSIG /".

http://codereview.chromium.org/126015/diff/5/1006#newcode44
Line 44: 
nit: blank line at EOF doesn't contribute anything positive

http://codereview.chromium.org/126015/diff/5/1007
File build/linux/dump_signature.py (right):

http://codereview.chromium.org/126015/diff/5/1007#newcode19
Line 19: sys.stderr.write("Error, did not read first page of data.\n");
Don't we want sys.exit(1) here too?

http://codereview.chromium.org/126015/diff/5/1007#newcode26
Line 26: out = ''
I have a better recipe for this, and it's endian-neutral:

# at top
import struct

# here
out = '%08x%04x%04x%02x%02x%02x%02x%02x%02x%02x%02x0' %
struct.unpack('I2H8B',struct.pack('16B',*signature))

(note I've left the 0 on the end.)

http://codereview.chromium.org/126015/diff/5/1007#newcode30
Line 30: out += '0'
I know what the 0 means, but will others?

Put a comment here, even just a same-line deal like

out += '0'  # Linux always uses generation 0