761907 - Ensure ARM crash dump generation works in gonk

Status: RESOLVED FIXED

(Toolkit :: Crash Reporting, defect)

Description

[Chris Jones [:cjones] inactive; ni?/f?/r? if you need me]

The gonks we care about are just ARM/bionic, so the existing breakpad code should work out of the box.  But we need to verify that.

Comment 1

[(not currently active) Ted Mielczarek]

I ran b2g on my SGS2 and was able to `kill -ABRT` the b2g process and get a minidump out. I need to track down one issue though: the module debug identifiers are coming out all zeroed out.

Comment 2

[(not currently active) Ted Mielczarek]

Okay, I figured out the root cause there, I have a couple of Breakpad patches.

The first is:
http://codereview.appspot.com/6349099/

Comment 3

[(not currently active) Ted Mielczarek]

The second is:
http://breakpad.appspot.com/414002

Comment 4

[(not currently active) Ted Mielczarek]

With those patches, looks good:
31  b2g!main [nsBrowserApp.cpp : 153 + 0xf]
     r4 = 0x404f3641    r5 = 0x00000000    r6 = 0x00000001    r7 = 0xbebb7bf4
     r8 = 0x00000000    r9 = 0x00000000   r10 = 0x00000000    fp = 0x00000000
     sp = 0xbebb5b18    pc = 0x0000896b
    Found by: call frame info

Comment 5

[(not currently active) Ted Mielczarek]

Attachment: Breakpad tweaks to make things work properly on gonk

There was a weird bug in Breakpad's use of linux-syscall-support. We were using sys_stat (not sys_stat64) because of some NDK #defines, but that syscall apparently doesn't work in some situations (ICS? new kernel? not sure exactly). These patches have upstream review, so I'm just going to land them when I get a chance.

Comment 6

[(not currently active) Ted Mielczarek]

Pushed the patches that got r+ upstream:
http://hg.mozilla.org/integration/mozilla-inbound/rev/a4a0f1412b9d

Comment 7

[Ed Morley [:emorley]]

https://hg.mozilla.org/mozilla-central/rev/a4a0f1412b9d