Hi Sam. So it was YOU who repaired the JMeter build on the 2nd Jan! I was about to go mad. Anyway, thanks.
I think I get the picture now. Would you (or someone else) be so kind as reviewing this for severe misunderstandings?
* Gump reads the project descriptors to establish a dependency tree, then builds each project in order.
* Gump uses a modified "ant" which totally ignores all "classpath" settings -- instead it places all JARs exported by projects depended upon into the classpath.
Q1: is it then best practice to remove all those jars from our lib/*.jar in CVS?
* Thus the Gump nightly builds always happen against the CVS versions of all projects -- which means that if I build JMeter from the sources on my workstation I won't get the exact same results (since I will be building against released versions of those projects).
* The copying of build results to http://cvs.apache.org/builds/jakarta-jmeter/nightly/ is not done by Gump. (Even though it must be run just after (some?) builds).
Q: how can I have the JMeter distribution files published there?
* The publishing of the src snapshot in that same directory has nothing to do with Gump.
Salut,
Jordi.
Sam Ruby wrote:
Jordi Salvat i Alabart wrote:Hi Sam. I'm taking this out of general to avoid bugging everyone around with my stupidness.
Don't be afraid to post e-mails like these on jakarta-general. That way everybody can help, and everyone can learn from the answers.
First of all, thanks again for your help, and my apologies if you just don't have time to help newcomers.
I do get a lot of e-mails, so I may occasionally miss one. If that happens, try again. Or, ask on general.
I have many questions about the nightly build & snapshot process, how it names the results and where it places them. Every project in Jakarta seems to work in a slightly different way -- and JMeter most different of all. I also have questions on how the web site is generated and published (I'd like JMeter's to be generated from the xdocs instead of just pulled out from cvs/jakarta-jmeter/docs).
There are some general policies, and then each subproject in Jakarta is free to tailor them to their own needs.
The nightly build simply checks out jakarta-jmeter and builds the default target. It is driven from the following project definition:
http://cvs.apache.org/viewcvs/jakarta-gump/project/jakarta-jmeter.xml
I am willing to upload to the jakarta web site anything that build produces, simply ask.
The way most jakarta projects work, as does jmeter, is that there is a "docs" target in the build that generates the docs directory from the xdocs sources. Simply build this and then commit the generated docs. Then somebody needs to go onto the daedalus.apache.org and execute a cvs update to get resynch the website with the cvs tree. For a number of projects, I have a script that automatically does this four times a day, so this eliminates a step... all you need to do is build and commit. Jmeter is not currently in this script... but can be added upon request.
I've been checking the FAQ at http://www.apache.org/dev/committers.html#web, but it didn't help much. Either it's obsolete, or it doesn't apply to all jakarta projects.
Is there any documentation I could read to get better informed?
Most of the jakarta documentation is on jakarta.apache.org. Try clicking on "gump" or "Website Maintenance".
- Sam Ruby
-- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>
