[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [gpsd-dev] Jon's commit 70e9681
From: |
Fred Wright |
Subject: |
Re: [gpsd-dev] Jon's commit 70e9681 |
Date: |
Fri, 4 Mar 2016 11:29:26 -0800 (PST) |
On Thu, 3 Mar 2016, Jon Schlueter wrote:
> On Thu, Mar 3, 2016 at 7:26 PM, Fred Wright <address@hidden> wrote:
[...]
> > I only just noticed that in the midst of this discussion (I believe before
> > I posted the above), commit 70e9681 added 'testprogs' to the 'check'
> > target. It also moved 'testclean' to the front, which is inconsistent
> > with the intent of commit 9517cc8 ("...and have the check target clean up
> > after itself.").
> >
> > Comments?
>
>
> your right. I didn't dig back through the history like I should have.
>
> Change reverted and pushed
We seem to be almost going in circles here. :-) Commit c26f063 undid
commit 70e9681, but then commit f402310 removed 'testclean' from check
altogether, with the comments:
as there is a report of testclean running before all
of the tests run and causing a failure
I remember there is some ordering issues with scons and
multiple targets or multiple threads and ordering
This was always known, which is why I added the 'test-noclean' target.
But if it's generally agreed that it's preferable for 'check' not to
include 'testclean', then I might as well remove the redundant
'test-noclean'. The 'check' target already has two names; it doesn't
really need three. :-)
Ordinarily, I'd be reluctant to remove a user-visible feature for fear of
breaking something, but since 'test-noclean' has only existed for a few
days, that's probably not an issue.
Meanwhile, anyone wanting the old behavior can always do:
scons check testclean
Or for parallel testing without conflicts:
scons check -jNN check && scons testclean
Fred Wright