On 2/7/07, Gregory Shimansky <[EMAIL PROTECTED]> wrote:
Geir Magnusson Jr. wrote:
>
> On Feb 6, 2007, at 5:48 PM, Gregory Shimansky wrote:
>
>> Geir, such messages are counter productive.
>
> It's not counter productive, just incomplete :)
>
>> Could you at least write on what platform you're running the tests?
>> I've just ran the tests on my bleeding edge Gentoo installation on x86
>> and they all passed. So no, I don't see drlvm tests failing.
>
> I'm just watching my machines (ubuntu x86) and getting failures. I'm
> testing again to narrow it down. I was also watching the alerts stream
> and we had by my count 10 failure messages today...
Speaking of which I was going to ask this question a long time ago. I am
not understanding these alerts at all.
I am seeing a message like this (skipping unnecessary whitespace):
[build-test alert] BUILD FAILED lnx gcc debug: drlvm
Errors/Warnings: (1)
Unit Tests: (2310)
All Tests Passed
and wonder why the preceeding message like this
[build-test alert] BUILD SUCCEEDED lnx gcc debug: classlib over drlvm
Errors/Warnings: (169)
Unit Tests: (19455)
All Tests Passed
was actually successful when there were 169 errors/warnings and in the
first it was only 1.
Gregory, let me give some explanation (by the way, it will good if you
ask direct question when you need some clarification :) ).
First of all, the number of errors/ warnings is calculated as number
of the records with error/warning level in the log. Initially these
messages were removed from notifications due to lots of warnings
(notification size was ~3M). Now these messages returns back (harmony
3060). Also, failed cunit/ smoke tests name for DRLVM was added to
notifications (harmony-3013).
I have one big issue to resolve: run drlvm build as 'ant' instead of
cmd/sh process to manage execution log in common manner.
That is for now I continue send explanation manually about my
notifications for DRLVM failures if it not obvious from the
notification what should be fixed.
Thanks, Vladimir
These alerts contain so little information that it is impossible to
understand if it is a bug in the scripts (which I tend to think so far)
or something really is broken.
--
Gregory
