[gmf-dev] GMF 2.1.0M6 :: Status Update

[Nick Boldt <codeslave@xxxxxxxxxx>]

Thanks to Max applying my patches on the weekend and this morning, and to
Anthony's map & tag fixes, the GMF builds are now running, tests included.
The full run takes about 33 mins.

There are however a number of test failures and errors. As such, I have
some questions about how to proceed with tonight's build. See below for
details.

-------------

As we have this build system set up on two servers now, one with Ubuntu
6.06LTS and one with Fedora Core 5 (Xen), I've been running on both to see
if any of these problems are server-specific (kernel, memory, SMP, GTK).

Here are the results:

http://emf.torolab.ibm.com/modeling/gmf/downloads/?showAll=1&hlbuild=N200804071649&sortBy=date&project=gmf#N200804071649
 (Ubuntu: 2 errors, 9 failures)
http://emft.eclipse.org/modeling/gmf/downloads/?showAll=1&hlbuild=I200804071746&sortBy=date&project=gmf#I200804071746
 (FC: in progress)

Because this is the first time we've had the tests running, we've not yet
tested the promote system to ensure that it will create all the metadata
(Ganymede .sc file, update site, RSS feed, releases notes) as expected --
the first time is always interesting. And because of time zone offsets, I
won't necessarily be available at the same time as Max tonight/tomorrow in
order to test out the system and fix any glitches that might arise should
he run it while I'm asleep.

So there are the questions I have for tonight's M6 release.

1. Is it acceptable to publish a build w/ JUnit test failures and errors?
Assuming we publish w/ failures/errors, do you want those hidden from the
downloads page, suggesting the build is green until someone digs down into
the test results to see the actual problems? Or, would it be more
acceptable to disable those tests so that only the ones we know are good
will run, then re-enable the others on Wednesday?

2. Because we have emf.toro and emft.eclipse both set up to run GMF builds
using the exact same build system, we can this week publish from the
emf.toro server rather than the emft.eclipse server. This will allow
Anthony and myself to be online concurrently tonight and address any
problems that arise without having to deal with timezone delays. Is that
acceptable?

3. If either (1) or (2) are not acceptable, is it acceptable to delay
publishing the M6 build until such time as Max & I are both online
tomorrow, eg., 10am EST? Is it better to be late, or to have errors?

--
Nick Boldt :: Release Engineer, IBM Toronto Lab
Eclipse Modeling :: http://www.eclipse.org/modeling
http://wiki.eclipse.org/index.php/User:Nickb