Hudson is failing on trunk

[robe]

The failure is a mystery as everything gets built, but it seems to be trying to send email and fails on send.

Sending e-mails to: info@… ... Notifying upstream projects of job completion Finished: FAILURE

This has been happening since early r8723

[chodgson]

If you go and look at hudson there are currently two builds failing:

​http://office.refractions.net:1500/

regress-1.5 is failing (for the past month) because of tickets test #1344 producing different results due to changes in the hex output format in postgres 9.0... not sure how best to address that.

trunk is failing because of a XML error apparently introduced in r8735. The error message you quoted appears at the end of the failed trunk build, after a whole bunch of xml errors: ​http://office.refractions.net:1500/job/PostGIS-trunk-build/lastFailedBuild/console

Note that all this information is available from the the Hudson interface, with no login required, only a minute or two of poking around.

[chodgson]

Actually looking at the revisions, it was r8734 that introduced it, you are missing some </listitem> in the area you fixed the typos in for r8735.

Hudson waits for a few minutes before building to make sure there aren't more commits coming, and 8735 was committed very shortly after 8734.

[chodgson]

Sorry I fail at reading commit logs. Reading the error message the problem is actually with <comand> instead of <command>.

[robe]

okay have fixed that one at r8754. Though it was failing before my commits or did you fix those issues?

[chodgson]

Looking at the output of the latest hudson trunk build:

​http://office.refractions.net:1500/job/PostGIS-trunk-build/3648/console

There appear to be some other docbook/xml issues. Do you not get those errors when you build locally? Or are they expected?

[robe]

I've always seen them on Hudson and strk has complained about those, but Hudson has never had a problem with it and has succeeded even with them. so I don't think that is the core of the problem why it is registering a failure.

To be honest I am not setup to build the pdf so don't know.

If you look at: ​http://office.refractions.net:1500/job/PostGIS-trunk-build/3609/console

Which was a successful build, it has those notices too.

[robe]

Chris,

Any luck with this. Like I said I don't think the document building is the culprit because it builds the PDF successfully. What I do notice comparing the last success:

​http://office.refractions.net:1500/job/PostGIS-trunk-build/3609/

With the failures is that the failed ones don't have these lines after successful build of the PDF:

+ rm -f postgis.xml
+ mv postgis.xml.orig postgis.xml
++ grep Error doc-errors.log
++ wc -l
+ '[' 0 -gt 0 ']'
+ cd /var/home/buildbot/opt/hudson/jobs/PostGIS-trunk-build/workspace
+ /bin/cp build/doc/postgis-2.0.0SVN.pdf rel/doc
+ /bin/cp -r build/doc/html rel/doc
+ mv rel postgis-2.0.0SVN
+ tar cfz postgis-2.0.0SVN.tar.gz postgis-2.0.0SVN
+ echo SVN_REVISION=8715
+ echo POSTGIS_MAJOR_VERSION=2
+ echo POSTGIS_MINOR_VERSION=0
+ echo POSTGIS_MICRO_VERSION=0SVN

So it seems to be something is going wrong when its trying to clear out files. The failed ones just culminate in a after the successful PDF build.

+ exit 1

[robe]

Looking closer I think that is the packaging step when it packages and uploads

[chodgson]

The "exit 1" is coming from the test grepping the output of the doc build for "error" - guess what, it finds lots of errors mentioned, even though it eventually finishes with "successful build" - so I changed that check to grep for "successful build" instead, and only "exit 1" if it doesn't find it.

We now have a happy Hudson build of trunk!