Re: fixed.gnome.org deployment update

[Frederic Peters <fpeters 0d be>]

Iago Toral Quiroga wrote:

> BTW, I realized that the main page, which shows the summary for all
> projects, has one little old problem.  You can see that Cairo is shown
> as successfully built, however it has failed checks. The reason for
> this, AFAICS, is that the summary shows the result of the last step of
> the build. In the case of Cairo, there are failed checks, however, this
> is not considered a build failure and then we run the code coverage
> reports, which run ok. IMHO this is ok, because even if we have failed
> checks we still want to know about the code coverage, however, this
> makes the build summary show a "success" (because the last step was
> successfully executed) for Cairo instead of the more meaningful "Checks
> failed". 
>
> I guess we should add support for defining if a step is an essential
> part of the build (like checkout, make or make check) or if it is an
> auxiliary step (like the coverage reports), so that when we show the
> summary, we show the result of the last essential step. Just a quick
> thought...

Is this possible for the coverage reports to fail?  Because, yes, that
situation shouldn't be considered a build failure (for all but slaves
administators I guess).

I didn't check but it may just amounts to add

 def evaluateCommand(self, cmd):
     if cmd.rc != 0:
         return WARNINGS
     return SUCCESS

to the JHBuildModulePathCommand used for module-reports.sh

Then some little changes to ProjectsSummary.body (in buildbot/
status/web/__init__.py) to get id of "steps[-1]" and use
"get_check_step(steps)" instead, and we would be done.

I don't have coverage reports running, so I can't check this for sure,
but I'll try to patch jhbuild somehow this week-end.

Btw, do you have any uncommited jhbuild changes ?  I could certainly
look at them at the same time.


        Frederic