[postgis-devel] 3.1.0alpha2 test report

[Greg Troxel]

rmrodriguez at carto.com writes:

> Hi,
>
> On Wed, Jul 29, 2020 at 12:08 AM Greg Troxel <gdt at lexort.com> wrote:
>> Compared to 3.0.1, other than the usual bunch of upgrade scripts from/to
>> new versions, the set of installed files changed little, with two
>> withdrawals:
>
> Yes, those were removed on purpose (https://trac.osgeo.org/postgis/ticket/4579).

Thanks, good to know.  I like to chase down unexpected changes; they
turn out intended only 99% of the time.

>> It built fine.  Tests crashed on me, testing the not-yet-installed
>> package, and then I installed it and they still crashed the same way.
>>
>> I must have run tests on 3.0.1 when I updated, or I would not have
>> committed it.  But rebuilding 3.0.1 and running tests now, it crashes.
>> I will look into this; it looks like a proj problem.
>
> I agree that it looks like a proj issue. Its API changed in version
> 6+, so I'd try to make sure it's doing a clean build and compiling and
> linking against the same proj release (if you have multiple of them
> installed).

I only have one proj installed, and have been doing rebuilds of packages
in topological sort order relative to dependencies.  Things don't make
sense to me because my proj update was well before the postgis update:

        Updated geography/proj to 6.3.1 [gdt 2020-03-10]
        Updated geography/py-proj to 2.6.0 [gdt 2020-03-18]
        Updated geography/proj to 6.3.2 [gdt 2020-05-03]
        Updated geography/py-proj to 2.6.1 [gdt 2020-05-04]
        Updated databases/postgresql-postgis2 to 3.0.1 [gdt 2020-06-22]

and I always run the tests before pushing an update to pkgsrc.  I am
doing a rebuild of all proj/pgsql/postgis and will nuke my database
files and start fresh.

I will have to rebuild with debugging symbols so I can see what's going
on in proj when it is calling into libstc++.  I'll also check netbsd-9
which uses gcc 7 instead of gcc 5.