689595 - update CI build code to work with BB or Jenkins

Status: RESOLVED WONTFIX

(Tamarin Graveyard :: Build Config, defect, P3)

Description

[Trevor Baker]

A few problems when using compile-generic.sh from within Jenkins 

1. the script creates a builds dir outside of the repo, which is causing problems because is outside of the "workspace" in Jenkins 
2. I don't want to copy the builds to asteam by default. If I want to archive artifacts I can do it via an artifact publisher
3. nor do I want to delete the objdir

Comment 1

[Trevor Baker]

Attachment: condition code using JENKINS_HOME env variable

Comment 2

[Trevor Baker]

Attachment: condition code using JENKINS_HOME env variable

Comment 3

[Trevor Baker]

Attachment: condition code using JENKINS_HOME env variable

Comment 4

[Trevor Baker]

Attachment: condition code using JENKINS_HOME env variable

this passed in the sandbox

Comment 5

[Trevor Baker]

Attachment: patch to get compile and test working on mac and windows in jenkins

Comment 6

[Dan Schaffer]

Comment on attachment 563477 [details] [diff] [review]
patch to get compile and test working on mac and windows in jenkins

in compile-generic.sh did you mean to remove the last line rm -fr objdir  

I'm not convinced pre-pending the existing INCLUDE, LIB, LIBPATH values is good since it may create confusing situations.   Like we should either require the values to be set correctly on the machine and we assume they are already set or require the script to set the values.  Combining the two is confusing.

Comment 7

[Trevor Baker]

Attachment: patch to return integer rev id as environment variable

patch applies to:
https://github.com/jenkinsci/mercurial-plugin

commit 385b155c2366cc22f479b5288183bf853022dcee
Author: Jesse Glick <[e-mail]>
Date:   Fri Sep 9 16:09:38 2011 -0400

Comment 8

[Trevor Baker]

(In reply to Dan Schaffer from comment #6)
> in compile-generic.sh did you mean to remove the last line rm -fr objdir  

Yes, I am archiving the artifact from there are reusing it in downstream jobs. This is to work around the BB system putting the shells in a /builds/ dir outside the repo.  In Jenkins, I can't get outside of WORKSPACE, so addressing the shell from objdir is the intent.


> I'm not convinced pre-pending the existing INCLUDE, LIB, LIBPATH values is
> good since it may create confusing situations.   Like we should either
> require the values to be set correctly on the machine and we assume they are
> already set or require the script to set the values.  Combining the two is
> confusing.

At the heart of it is that jenkins is really flexible in letting you add and and remove nodes at will to scale up or down.  Most/all of our buildbot config is in hg, while it won't be practical to use the same mechanisms for Jenkins.  My thoughts on best practices for Jenkins are still formulating, but artificially constraining ourselves with such a flexible tool seems like the wrong approach.

Comment 9

[Dan Schaffer]

Attachment: patch to performance runtests fix, add socket timeout

Comment 10

[Trevor Baker]

Attachment: patch from HM repo

consolidated patch to bring TR up to date with HM build changes

Comment 11

[Tamarin Bot]

changeset: 6832:7cc3acef01d7
user:      Brent Baker <[e-mail]>
summary:   Update to latest jenkins tweaks based on bz:689595

http://hg.mozilla.org/tamarin-redux/rev/7cc3acef01d7

Comment 12

[Brent Baker]

Comment on attachment 571009 [details] [diff] [review]
patch from HM repo

removing review. Not sure how much of this is still to land. Most of it has already landed in tamarin-redux via the float merge