[Buildbot-devel] Hierarchical tests of different software sources, where some tests are allowed to fail

[tom fogal]

mobile, forgive my top posting + terseness.

i do this with a scheduler per project (i.e. one for boost, another for
A) and the latter projects use dependent schedulers.  I would put all
your 'make check' steps in the same builder that builds your software.

also take some time looking at 'haltOnFailure' and 'flunkOnFailure' in
the docs.

-tom

Peter Schueller <schueller.p at gmail.com> writes:
> Dear all,
> 
> I would like to setup the following build infrastructure with buildbot:
> 
> [multiple compilers used|multiple boost branches used]
> 
> 1) checkout boost and build it (exit if this fails)
> 2) checkout some other software A, build it (exit if this fails)
> 2.1) make check software A (continue if this fails)
> 2.2) make distcheck software A (continue if this fails)
> 3) install software A into a specific directory (exit if this fails)
> for each plugin B,C,D,... that depends on software A:
> 4.B) checkout plugin B, build it using installed software A (exit for
> branch 4.B if this fails)
> 4.B.1) make check plugin B (continue if this fails)
> 4.B.2) make distcheck plugin B (continue if this fails)
> 4.C) checkout plugin C, build it using installed software A (exit for
> branch 4.C if this fails)
> ...
> 
> So I need a possibility to do a hierarchically scheduled build, where
> steps at the "leaves of the tree" are noncritical for subsequent
> steps, and steps at "inner nodes of the tree" are critical for
> subtrees.
> 
> I found the possibility of the "Dependent Scheduler" but the
> documentation tells me that there is no guarantee this scheduler will
> use the correct revision number (e.g., the same one for steps 4.B and
> 1).
> 
> The same documentation tells me that with the Triggerable Scheduler
> this should be easier or better possible, but I do not see how I can
> do noncritical steps using that scheduler, or critical ones (if the
> default is noncritical).
> 
> Furthermore it is not clear to me whether I could setup the whole
> thing even better, namely as follows:
> * some place to build the most recent boost branches to
> * building boost is triggered nightly and by commits
> * some place to build the most current version of A
> * building and testing A is triggered nightly and by commits
> * some place to build each plugin B,C,D
> * building and testing B,C,D is triggered independently nightly and by commit
> s
> * builds of B,C,D always use the most recent successful build of A
> 
> I would really appreciate some hints which strategy to adopt and where
> to look for more information (pointers to documentation and sourcecode
> are fine).
> 
> Best,
> Peter
> 
> -----------------------------------------------------------------------------
> -
> Learn how Oracle Real Application Clusters (RAC) One Node allows customers
> to consolidate database storage, standardize their database environment, and,
>  
> should the need arise, upgrade to a full multi-node Oracle RAC database 
> without downtime or disruption
> http://p.sf.net/sfu/oracle-sfdevnl
> _______________________________________________
> Buildbot-devel mailing list
> Buildbot-devel at lists.sourceforge.net
> https://lists.sourceforge.net/lists/listinfo/buildbot-devel