Enable coredumps on arm, assert for assumptions about sysctl coredump settings

Enable coredumps on arm, assert for assumptions about sysctl coredump settings

R=ricow@google.com

Committed: https://github.com/dart-lang/fletch/commit/97dae439cd1d8700738f442206a9d8c059f46531

[kustermann, 2015-08-25 09:08:24]

kustermann@google.com changed reviewers:
+ ricow@google.com

[kustermann, 2015-08-25 09:08:25]

https://codereview.chromium.org/1316663003/diff/1/tools/bots/fletch.py
File tools/bots/fletch.py (right):

https://codereview.chromium.org/1316663003/diff/1/tools/bots/fletch.py#newcod...
tools/bots/fletch.py:418: assert core_pattern == 'core'
I'm aware that the pattern could be made to include pid and more, but the Vms
are configured this way ATM.

[ricow1, 2015-08-25 09:10:09]

lgtm

https://codereview.chromium.org/1316663003/diff/1/tools/bots/fletch.py
File tools/bots/fletch.py (right):

https://codereview.chromium.org/1316663003/diff/1/tools/bots/fletch.py#newcod...
tools/bots/fletch.py:419: assert core_pattern_uses_pid == '1'
should we gracefully fail here, and just tell people that coredumps is not
available until the pattern is set?

[kustermann, 2015-08-25 11:14:48]

Thanks, I'll land it now with the assertion, I think it's really important for
us to now when coredumps can't be archived.

https://codereview.chromium.org/1316663003/diff/1/tools/bots/fletch.py
File tools/bots/fletch.py (right):

https://codereview.chromium.org/1316663003/diff/1/tools/bots/fletch.py#newcod...
tools/bots/fletch.py:418: assert core_pattern == 'core'
On 2015/08/25 09:08:25, kustermann wrote:
> I'm aware that the pattern could be made to include pid and more, but the Vms
> are configured this way ATM.

Also account for the newline in the file now.

https://codereview.chromium.org/1316663003/diff/1/tools/bots/fletch.py#newcod...
tools/bots/fletch.py:419: assert core_pattern_uses_pid == '1'
On 2015/08/25 09:10:09, ricow1 wrote:
> should we gracefully fail here, and just tell people that coredumps is not
> available until the pattern is set?

I thought about it as well, but
* users are not running this code, only bots are
* if a bot doesn't have it enabled, we really need to know and change that

That's why I decided against it so far.

[If we add a new VM it can easily happen it doesn't have the bit, and we'll
never notice, which is bad.]

[kustermann, 2015-08-25 11:15:54]

Committed patchset #2 (id:20001) manually as
97dae439cd1d8700738f442206a9d8c059f46531 (presubmit successful).