[Buildroot] google-breakpad experiences?

[Buildroot] google-breakpad experiences?

[Thomas De Schampheleire]

Hi,

Has anyone successfully used google-breakpad on MIPS64 and/or PowerPC targets?

For MIPS64/n32 ABI I had to make a few compilation fixes, but a simple
test program (copied from their wiki) does not provide any dump.

For PowerPC I added some code to make it compile, and although the
support I added is not yet complete, I could build it but it does not
produce any dump either.

On the MIPS version I did some debugging and found that the dumper
child was created but did not seem to run, and thus no dump was taken.

When building for x86_64 (qemu) things worked fine.

Did any of you succeed in using google-breakpad on embedded targets?

Thanks,
Thomas

[Buildroot] google-breakpad experiences?

[Will Wagner]

On 15/10/2014 15:53, Thomas De Schampheleire wrote:
> Hi,

<snip>

>
> Did any of you succeed in using google-breakpad on embedded targets?

Hi Thomas,

We have successfully been using google-breakpad for a few years. However 
it is only on x86 and arm targets and we have our own internal package 
(it pre-dates the one now upstream but is broadly similar, we have a 
number of custom patches as well).

Are you trying to generate dumps within the crashed process or are you 
running the crash server? We have found the latter to be much more reliable.

Regards
Will


-- 
------------------------------------------------------------------------
Will Wagner                                     will_wagner at carallon.com
Development Manager                      Office Tel: +44 (0)20 7371 2032
Carallon Ltd, Studio G20, Shepherds Building, Rockley Rd, London W14 0DA
------------------------------------------------------------------------

-- 
------------------------------------------------------------------------
Will Wagner                                     will_wagner at carallon.com
Development Manager                      Office Tel: +44 (0)20 7371 2032
Carallon Ltd, Studio G20, Shepherds Building, Rockley Rd, London W14 0DA
------------------------------------------------------------------------

[Buildroot] google-breakpad experiences?

[Thomas De Schampheleire]

Hi Will,

On Wed, Oct 15, 2014 at 5:05 PM, Will Wagner <willw@carallon.com> wrote:

> We have successfully been using google-breakpad for a few years. However it
> is only on x86 and arm targets and we have our own internal package (it
> pre-dates the one now upstream but is broadly similar, we have a number of
> custom patches as well).
>
> Are you trying to generate dumps within the crashed process or are you
> running the crash server? We have found the latter to be much more reliable.

Currently I'm simply trying the test program described at:
https://code.google.com/p/google-breakpad/wiki/LinuxStarterGuide

which is dumping from within the crashed process.

How does this crash server work? I must say the documentation of
google-breakpad is not very extensive.

Could you also elaborate on the unreliabilities you have encountered
with the first method?

Thanks,
Thomas

[Buildroot] google-breakpad experiences?

[Will Wagner]

On 15/10/2014 19:44, Thomas De Schampheleire wrote:
> Hi Will,
>
> On Wed, Oct 15, 2014 at 5:05 PM, Will Wagner <willw@carallon.com> wrote:
>
>> We have successfully been using google-breakpad for a few years. However it
>> is only on x86 and arm targets and we have our own internal package (it
>> pre-dates the one now upstream but is broadly similar, we have a number of
>> custom patches as well).
>>
>> Are you trying to generate dumps within the crashed process or are you
>> running the crash server? We have found the latter to be much more reliable.
>
> Currently I'm simply trying the test program described at:
> https://code.google.com/p/google-breakpad/wiki/LinuxStarterGuide
>
> which is dumping from within the crashed process.

The example works fine on our platforms.

>
> How does this crash server work? I must say the documentation of
> google-breakpad is not very extensive.

You start up another process before creating the ExceptionHandler object 
(passing in a descriptor via server_fd). The advantage of out-of-process 
dump generation is that you don't have to worry about a corrupted 
environment & C library.

The documentation is indeed poor and help on the forums was never great 
either. I tried to upstream our patches but never got any response on 
them either.

>
> Could you also elaborate on the unreliabilities you have encountered
> with the first method?

The problem we have seen with this is that in a number of crash 
scenarios it never produces a dump because something is wrong in the c 
library. Also we are able to put extra functionality into the crash 
server process so it can offer to email files, save to usb key etc.


-- 
------------------------------------------------------------------------
Will Wagner                                     will_wagner at carallon.com
Development Manager                      Office Tel: +44 (0)20 7471 9224
Carallon Ltd, Studio G20, Shepherds Building, Rockley Rd, London W14 0DA
------------------------------------------------------------------------