Nightly failure

[Bengt Rutisson]

Hi Jesper,

I assume that you are talking about nsk/regression/b4212983.

In that case I agree that the test is bad. It will not detect if we 
start allowing -Xmx0m since it only checks the error message for the 
initial heap size.

It is probably too late by now, but I would suggest filing a bug that 
matches the test failure to make this a know failure. The bug should be 
on hotspot/test and say that this test needs to be split up into two 
test cases. One for -Xms0m and -Xmx0m.

If you do that I think you can push hotspot-gc to main. But maybe we 
have already missed the snapshot deadline?

Bengt



On 5/9/13 12:41 PM, Jesper Wilhelmsson wrote:
> Hi,
>
> There is a failure in the nightlies due to change 4612 where the 
> argument parsing was changed somewhat.
>
> The test starts a JVM with invalid heap size flags and looks for a 
> specific error message. However, since there are more than one invalid 
> flag on the command line (-Xms0m -Xmx0m) we now check the flags in a 
> different order than before and a different error message is printed 
> causing the test to fail.
>
> I find this test extremely unstable even though it is in our "stable" 
> GC baseline. If we want to check for invalid arguments, there should 
> only be one invalid argument on the command line.
>
> I'm waiting for a response from SQE on this and in the mean time I 
> will not push hotspot-gc to main. I don't feel comfortable enough 
> around mercurial to attempt to exclude this changeset from the push. 
> Especially since there is nothing wrong in the change.
>
> I have a job in the jprt queue now to pull main to gc.
>
> Let me know if there is something in hotspot-gc that needs to be 
> pushed to main this week.
> /Jesper

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://mail.openjdk.org/pipermail/hotspot-gc-dev/attachments/20130510/c08c307c/attachment.htm>