[nanomsg] Re: New fork (was Re: Moving forward)

  • From: Garrett D'Amore <garrett@xxxxxxxxxx>
  • To: Dirkjan Ochtman <dirkjan@xxxxxxxxxx>, martin@xxxxxxxxxx
  • Date: Fri, 20 Mar 2015 08:01:03 -0700

> On Mar 20, 2015, at 5:30 AM, Dirkjan Ochtman <dirkjan@xxxxxxxxxx> wrote:
> 
> On Fri, Mar 20, 2015 at 9:34 AM, Martin Lucina <martin@xxxxxxxxxx> wrote:
>> I was planning on migrating at least some of the slaves (Linux, FreeBSD) to
>> Xen VMs on different physical hardware which I host elsewhere, however if
>> you are going to migrate to Travis then I'd rather just save myself doing
>> the work and switch it all off.
> 
> IIRC Travis only runs Linux and a trial of OS X, so it might still be
> worth it to preserve some of the other stuff. On other hand, maybe
> Garrett has another solution in mind for this.

Looking at what we have today, if we can keep the build slaves running for just 
a wee bit longer, I think it would be useful.  I would like to get travis CI up 
and running — at least part of the reason for this is so that other people can 
follow the recipes there and validate their own PRs.  (Plus, Travis has *very* 
nice PR support— you can pre-test your PRs before merging, and everyone can see 
those test results, including the PR submitter.)

I'll have to chat up the Travis people to ask about OS X support.  I suspect 
that FreeBSD is not going to be easy to come by, and that support for Solaris 
is going to be nigh impossible — particularly SPARC.   But it also sounds like 
you’re planning on retiring the SPARC servers you have?

That said, we’re going to miss CI for a lot of other platforms — people are 
running nanomsg on ARM systems, in iOS, on OpenBSD, NetBSD, HPUX, VxWorks (?), 
etc.  We can’t have CI for all of them at this point in time.  The loss of 
FreeBSD would be a shame, and I have personal reasons for wishing we’d have CI 
support for Solaris or illumos (esp. the latter), but again, I don’t think any 
of the CI providers have an interest.

Right now, its unclear to me that our current CI setup does anything more than 
verifying that stuff builds.  The self tests don’t seem to be built unless 
building with cmake on Windows.  That’s something that needs to be fixed.  
Frankly, I think there is more value in having the self-tests execute, even if 
*only* on Linux, then on ensuring we are still building on other platforms. 
(Not that I want to stop supporting those other platforms.)

The other thing I’d to do is get us using appveyor for Windows testing.  I’m 
most definitely *not* a Windows developer, but it seems that there is a lot of 
effort that has been expended to support that platform, and it would be nice if 
we could ensure we didn’t regress on it.  That becomes especially important as 
I don’t have any systems running windows within normal easy reach of me.

        - Garrett


Other related posts: