Reviewer needed - fix for IcedTea bug#637

[Dr Andrew John Hughes]

On 11:22 Thu 10 Mar     , Mark Wielaard wrote:
> On Thu, 2011-03-10 at 00:10 +0000, Dr Andrew John Hughes wrote:
> > On 23:09 Tue 08 Mar     , Mark Wielaard wrote:
> > > No, it has indeed two functions. It does the summary and finishes any
> > > stray processes from the jtregcheck target(s). I guess you didn't like
> > > my suggestion for the name, and Pavel's orginal name jtregcheck-finish
> > > was the better one. I don't think the functionality should be split. But
> > > they can be if you want.
> > > 
> > I think your name is preferable.  I just don't see the point in
> > splitting this off at all in the way that has been suggested.  All it
> > does is break things.
> 
> I see you don't like the new functionality, but it is useful for others
> who do a lot of testing. It is not meant to just break things obviously.
> Lets see if we can come to some agreement to get the new functionality
> in a way that is more acceptable then.
> 

That would be good.  I see this patch seems to have been committed before this
happened though.  I've now reverted the sections under discussion which break
make -j (the jtreg-summary addition).

> > > > The current jtregcheck target runs after all required checks have been
> > > > run, prints a summary and kills off any processes floating around.
> > > 
> > > That is what jtregcheck-summary does to. It is the last dependent target
> > > that gets run after all requested tests have been run. It is just that
> > > the target has been split to have the last step independently.
> > > 
> > 
> > For absolutely no reason as far as I can see.  What makes you think it
> > will run last?  I'm not aware of any rule that says that and didn't
> > see anything in checking the GNU make manual.  This _will_ break -j
> > as there is nothing to stop the summary being rule.  At present, the
> > dependents are all fulfilled then the body of the rule is run.  This
> > breaks that for no reason.
> 
> Aha, ok, I see your point. You are worried about breaking make check -j.
> I admit to not have thought of that possibility. But it would be cool if
> that worked, then you could run some of the testsuites in parallel.
> 

Well it worked before, so the proposed change regresses.

> So what we would ideally want is:
> a) individual make check targets for the different testsuites,
>    that can be rerun separately
>    - this is by the make jtreg_check="..." check
>      which ultimately translates into the check-* targets,
>      and since they are "phony" targets it is possible to
>      run them independently.

I thought this was by make check-x which always worked just fine.  I
don't see why we need two ways of doing the same thing.

I would like a way of disabling which tests are run by the main 'make
check' target for make distcheck purposes as the jdk tests never all
succeed and take forever.  I've added this in IcedTea7 already and it
needs bringing over.

> b) have an indication of the failure of any individual testsuite
>    - this is done by the new tests in the check-* targets
>      which make the check targets fail if there are failures or
>      errors in the testsuite.

I'm not sure that's the most helpful thing.  Surely you could determine
this from the end log anyway?

> c) cleanup stray test processes and generate a summary of the
>    testsuite results done up till that point.
>    - this makes make check -k necessary, so that the summary/finish
>      target is always run.

I don't like the idea of killing processes mid-test.  That should be
done at the end and only at the end.

> d) make all the above possible with -j so multiple testsuites can be
>    run in parallel.
>    - No solution yet?

That worked before this patch.

> e) make the cleanup/summary target run even without -k
>    - No solution yet?
> 
> Any suggestions how to achieve that in a cleaner way than is done now?
> 

I don't see a problem with how it was to start with.

> Thanks,
> 
> Mark
> 

-- 
Andrew :)

Free Java Software Engineer
Red Hat, Inc. (http://www.redhat.com)

Support Free Java!
Contribute to GNU Classpath and IcedTea
http://www.gnu.org/software/classpath
http://icedtea.classpath.org
PGP Key: F5862A37 (https://keys.indymedia.org/)
Fingerprint = EA30 D855 D50F 90CD F54D  0698 0713 C3ED F586 2A37