Re: [gpsd-dev] [gpsd-users] Warnings from 32 bit FreeBSD

[Greg Troxel]

"Gary E. Miller" <address@hidden> writes:

> Yo Greg!
>
> On Thu, 20 Jun 2019 10:13:02 -0400
> Greg Troxel <address@hidden> wrote:
>
>> "Gary E. Miller" <address@hidden> writes:
>> 
>> >> Avoiding the need for the GNU-specific --transform would be nice,
>> >> but not trivial.  
>> >
>> > There is really no need for "scons dist" to be run more than once,
>> > on just one OS, for every release.  So not high on my list.  
>> 
>> That's not really true.  Random packagers running dist to make a
>> tarball to test via the packaging system is important.
>
> I'm happy to provide things packagers use.  But that is a very small
> class of users.  And I fear the creation of official looking release
> tar balls that are not official.
>
> Is this only for package debug?

Yes; I would never distribute such a package.  It's to be able to do the

  source
  build release
  use release to build package
  install package
  test installed packaeg

path, which tests what actually happens when there is a release and
people use it.

Generally, projects that have even half-way regular releases (and gpsd
is in really good shape relative to that) do not have issues with cowboy
distributions being published.

But I agree that it's more work to fix than is worth it, and just left
better comments.

>> But we have a workaround for the problem of scons not being as good
>> autotools :-)
>
> Once again an unsubstantiated opinion. :-)

We all have those, in good fun.  In this case, it's not really an
opinion: scons is failing to produce a tarball because it doesn't
adequately deal with portability issues.  automake is known to build
tarballs pretty much everywhere.  (I realize there are multiple points
of comparison and this is just one of them.)