[wtp-releng] Some notes on infrastructure and builds this week

[David M Williams <david_williams@xxxxxxxxxx>]

Just to keep everyone informed of the
same information I know, as little as it is. 

_Some_ of the build failures and JUnit
failures earlier this week have been confirmed to be due to odd, not completely
understood timing issues on the Eclipse infrastructure. 
Not all these failures are visible,
but a few times, I've seen failures that halt the build, because something
couldn't be found/checked out of CVS. 
Then the next attempt would fail, at
a different point, then again at a different point, then finally be able
to pull out enough to complete. 
One reason it was more visible on the
JSF component is that both JSF and JPT have been set to _not_ halt the
build if they fail (since leaf components) 
whereas WST or JST will completely halt
the build, if part of it fails (since they are not leaf components), and
in the case of a complete halt, there is nothing
published to download page. 

What we _think_ is going on, is that
some CVS check-ins are not fully "replicating" throughout the
cluster right away (maybe taking 1 to 30 minutes?, or more?) 
instead of the usual "minute or
less".  So, we'll keep a closer eye on this, but not sure what
there is to do, except stop checking in so much code! :) 
I suspect mostly due to busy servers
due to everyone being busy towards release, and there was some recent DoD
attacks (or something) at Eclipse that 
might have chocked things up a bit.



Also, I've confirmed the recent notes
about "vservers" should not effect us in WTP, and this morning's
"firewall" problem in that one JEM test is currently a mystery

and probably transient.