CQ: don't give up on update and clobber build failures

CQ: don't give up on update and clobber build failures

With HEAD being possibly broken these should not be fatal.

BUG=none

Committed: http://src.chromium.org/viewvc/chrome?view=rev&revision=234892

[M-A Ruel, 2013-11-12 19:47:27]

This will somewhat increase the load. Since the CQ doesn't understand about
broken tests at HEAD either, I'm not sure how valuable this is.

So you want to have update failures to be retried? A compile failure is mostly
always a read problem is not worth being retried.

What are you trying to solve here?

[Paweł Hajdan Jr., 2013-11-12 20:13:15]

On 2013/11/12 19:47:27, M-A Ruel wrote:
> This will somewhat increase the load. Since the CQ doesn't understand about
> broken tests at HEAD either, I'm not sure how valuable this is.
> 
> So you want to have update failures to be retried? A compile failure is mostly
> always a read problem is not worth being retried.
> 
> What are you trying to solve here?

Please see e.g. https://codereview.chromium.org/61913005/ and
https://codereview.chromium.org/67223002/

Users just CQ changes again anyway.

[M-A Ruel, 2013-11-12 20:19:38]

On 2013/11/12 20:13:15, Paweł Hajdan Jr. wrote:
> On 2013/11/12 19:47:27, M-A Ruel wrote:
> > This will somewhat increase the load. Since the CQ doesn't understand about
> > broken tests at HEAD either, I'm not sure how valuable this is.
> > 
> > So you want to have update failures to be retried? A compile failure is
mostly
> > always a read problem is not worth being retried.
> > 
> > What are you trying to solve here?
> 
> Please see e.g. https://codereview.chromium.org/61913005/ and
> https://codereview.chromium.org/67223002/
> 
> Users just CQ changes again anyway.

Exactly, in both cases, the slave was dead, so *immediately* retrying wouldn't
have helped.

[Paweł Hajdan Jr., 2013-11-12 20:23:06]

On 2013/11/12 20:19:38, M-A Ruel wrote:
> On 2013/11/12 20:13:15, Paweł Hajdan Jr. wrote:
> Exactly, in both cases, the slave was dead, so *immediately* retrying wouldn't
> have helped.

Wouldn't the request go to another slave?

[M-A Ruel, 2013-11-12 20:28:09]

On 2013/11/12 20:23:06, Paweł Hajdan Jr. wrote:
> On 2013/11/12 20:19:38, M-A Ruel wrote:
> > On 2013/11/12 20:13:15, Paweł Hajdan Jr. wrote:
> > Exactly, in both cases, the slave was dead, so *immediately* retrying
wouldn't
> > have helped.
> 
> Wouldn't the request go to another slave?

Likely not, because this slave will stay at the top of the available slaves
list.

[Paweł Hajdan Jr., 2013-11-13 01:46:44]

CC Dirk

And now we have https://code.google.com/p/chromium/issues/detail?id=318520 - do
you want to close that as WontFix?

How about giving this CL a try - if it puts too much load on the TS we can
revert it.

[Dirk Pranke, 2013-11-13 01:56:17]

The problem I'd want to fix is to keep the CQ from giving up on patches due to
reasons having nothing to do w/ the patch itself. That just forces the user to
have to re-submit the patch to the CQ, and it's really annoying. 

It's also confusing, since the user is being told that the patch failed for said
unrelated reasons (and figuring that out is not immediately obvious to most
people).

If ToT is broken, then the CQ should use its normal logic for monitoring tree
closure and retrying jobs to work around it.

[M-A Ruel, 2013-11-13 02:06:07]

ok fine, I'll assume you'll monitor usage.

lgtm

[commit-bot: I haz the power, 2013-11-13 20:30:37]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/phajdan.jr@chromium.org/66963004/1

[commit-bot: I haz the power, 2013-11-13 20:30:51]

Change committed as 234892