[yocto] Multiple conncetion to svn (svn+ssh)

[Bryan Evenson]

Paul,

> -----Original Message-----
> From: Paul Eggleton [mailto:paul.eggleton at linux.intel.com]
> Sent: Monday, July 06, 2015 10:08 AM
> To: Marcin Krzemiński
> Cc: Bryan Evenson; yocto at yoctoproject.org
> Subject: Re: [yocto] Multiple conncetion to svn (svn+ssh)
> 
> On Monday 06 July 2015 12:57:39 Bryan Evenson wrote:
> > Marcin,
> >
> > > From: yocto-bounces at yoctoproject.org
> > > [mailto:yocto-bounces at yoctoproject.org] On Behalf Of Marcin
> > > Krzeminski
> > > Sent: Friday, July 03, 2015 7:55 AM
> > > To: yocto at yoctoproject.org
> > > Subject: [yocto] Multiple conncetion to svn (svn+ssh)
> > >
> > > Hello again,
> > >
> > > I have 12 recipes that download code from same svn repository (using
> > > svn+ssh)protocol. Recipes are grouped in packagegroup.
> > > When I want to bitbake packagegroup fetcher fail, but when I run
> > > recipe alone all is ok. I think it is because I want to open 8
> > > connection to one svn repository.How can I fix this, for example by
> > > allowing only eg. 2 recipes from packagegroup to be executed in paralell.
> > There is no fetcher-specific variable that I know of, but I think you
> > can set PARALLEL_MAKE="2" in your recipes to reduce the number of
> > fetches on your repository at a time.  The downside is your build will
> > go slower when it is building your recipes as it won't be doing as
> > many things in parallel.
> 
> That's not going to help. The parallelism we are talking about here is across
> recipes - PARALLEL_MAKE is just passed through to make within one task in
> one recipe, and make isn't even involved at the fetch stage.

Sorry, I grabbed the wrong variable.  I meant BB_NUMBER_THREADS, not PARALLEL_MAKE.

> 
> Something that might work would be to set a lockfile on the do_fetch task
> such that only one of the recipes could fetch at once. That could not allow
> two executing at once, but at least it would solve the problem. e.g. you could
> add this to all of the recipes:
> 
> do_fetch[lockfiles] += "${TMPDIR}/mysvnlock.lock"
> 
> (The file name isn't critical, it just needs to be the same for all of the recipes
> you wish to participate in the exclusive fetching.)

I had no idea that we could do this.  I don't see any documentation on lockfiles anywhere.
If you use a lockfile, do the all recipes with the same lockfile wait until the lockfile is available
before continuing on with that step?  Is there a timeout for waiting for the lockfile or does
the recipe wait indefinitely?

Regards,
Bryan

> 
> Cheers,
> Paul
> 
> --
> 
> Paul Eggleton
> Intel Open Source Technology Centre