Skip to main content

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index] [List Home]
Re: [virgo-dev] Hudson

Hi Hristo

Actually, the first thread below may indeed be a good clue for your problem and ours.

The Virgo build all-bundles macrodef in multi-bundle/common.xml might be to blame. This is used in various places to distribute a target across a number of bundles (projects). Here's the code:

	<macrodef name="all-bundles">
		<attribute name="target"/>
		<attribute name="buildpathRef" default="bundles"/>
		<element name="inherited-properties" optional="true" implicit="true"/>
		<sequential>
			<subant target="@{target}" verbose="true" buildpathRef="@{buildpathRef}">
				<inherited-properties/>
				<propertyset>
					<propertyref builtin="commandline"/>
				</propertyset>
				<property name="timestamp" value="${timestamp}"/>
			</subant>
		</sequential>
	</macrodef>

So we could look to replace the subant with an exec as in the thread below. This may solve your Ivy permgen problem as well as the compilation permgen problem we are seeing on Hudson.

Unfortunately, this is far from a simple cut'n'paste replacement, so if you have any ant experts who fancy coding a patch, we'd be very interested...

Regards,
Glyn

On 14 Oct 2010, at 10:00, Glyn Normington wrote:

> Thanks Hristo. The kernel build seems to be failing later than Ivy resolution now, so the Ivy/permgen issue is off our critical path for now. I hope you can make progress in the environments where this is an issue for you.
> 
> Regards,
> Glyn
> 
> On 14 Oct 2010, at 09:50, Iliev, Hristo wrote:
> 
>> Hi,
>> 
>> I searched the Ivy mailing list and found out that this problem is reported several times. The following threads suggest different workarounds:
>> 
>> ivy:buildlist/exec instead of subant:
>> http://www.mail-archive.com/ivy-user@xxxxxxxxxxxxxx/msg04370.html
>> 
>> loader attribute:
>> http://www.mail-archive.com/ivy-user@xxxxxxxxxxxxxx/msg00425.html
>> 
>> I think the problem with Virgo build is more related to the first issue.
>> 
>> Regards,
>> Hristo Iliev
>> 
>> 
>> -----Original Message-----
>> From: virgo-dev-bounces@xxxxxxxxxxx [mailto:virgo-dev-bounces@xxxxxxxxxxx] On Behalf Of Glyn Normington
>> Sent: Thursday, October 14, 2010 11:41 AM
>> To: Virgo Project
>> Subject: [virgo-dev] Hudson
>> 
>> After an extended and persistent investigation by Steve Powell and the Eclipse webmaster Denis Roy, the Hudson CI Virgo builds [1] are now back in good shape again.
>> 
>> All are building successfully except for the kernel which Steve is investigating.
>> 
>> The problem turned out to be a permgen size setting and may perhaps be related to the Ivy permgen issue that SAP mentioned on the community call. SAP agreed to raise the issue with the Ivy project and post a link here.
>> 
>> Regards,
>> Glyn
>> [1] https://hudson.eclipse.org/hudson/view/Virgo/
>> 
>> _______________________________________________
>> virgo-dev mailing list
>> virgo-dev@xxxxxxxxxxx
>> https://dev.eclipse.org/mailman/listinfo/virgo-dev
>> _______________________________________________
>> virgo-dev mailing list
>> virgo-dev@xxxxxxxxxxx
>> https://dev.eclipse.org/mailman/listinfo/virgo-dev
> 
> _______________________________________________
> virgo-dev mailing list
> virgo-dev@xxxxxxxxxxx
> https://dev.eclipse.org/mailman/listinfo/virgo-dev



Back to the top