[postgis-users] Compiling PostGIS 1.4.0 release on Solaris 10

[Paul Ramsey]

On Thu, Nov 26, 2009 at 7:29 AM, Nathan Widmyer <lighthousej at gmail.com> wrote:

> 1. I get differences in error message formats.
> The tests expect an error message, "ERROR:  parse error - invalid
> geometry" but the tests return "ERROR:  parse error - invalid geometry
> at character 18"
> I think the PGIS developers added "at character X" in a few error
> messages too, not just parse errors.
> Tests in error: regress, wkt, regress_ogc
> Possible Fix: update the tests to expect the extra information

That's extremely odd, any particular tarball should have tests and
expectations that match the version. Are you sure you don't have an
old PostGIS hanging around somewhere that doesn't match your test
suite?

> 2. The endianness of WKB that is expected differs from the actual WKB
> returned.  I have converted each WKB to WKT on both and they match
> exactly by their text representations.
> Tests in error: sql-mm-compoundcurve, sql-mm-curvepoly
> Possible Fix: have alternate expectation files for Solaris, compare
> the WKT instead of WKB, or if there's a way, force PGIS to write WKB
> in a particular endianness.

You're on an honest-to-god RiSC machine? nice. There's a second
parameter to AsBinary that you can use to control endianness. I think
we default to little endian, since that's so prevalent now, you if you
put 'xdr' into the second parameter you'll get out big endian.

P.

> Thanks,
> Nate
>
> On Wed, Nov 25, 2009 at 2:21 PM, Paul Ramsey <pramsey at cleverelephant.ca> wrote:
>> How old is your solaris? I just got my OpenSolaris VM back up and
>> running, but it's very fresh indeed. I'll get some testing going on
>> that.
>>
>> P
>>
>> On Wed, Nov 25, 2009 at 10:42 AM, Nathan Widmyer <lighthousej at gmail.com> wrote:
>>> Ah, thank you for clearing that up.
>>>
>>> I was going off of:
>>> http://postgis.refractions.net/documentation/manual-1.4/ch02.html#id2531756
>>>  which, respectively,
>>> Which is how I've always built software, test before install, so I
>>> didn't question the directions.
>>>
>>> I understand not conflicting with a running PostgreSQL, but is there a
>>> socket-only connection style like MySQL can do to connect to just that
>>> specific instance?
>>> I figured the tests were writing to a temp postgis.sql somewhere so it
>>> wouldn't require installing into PostgreSQL, and something like a sed
>>> or shell command wasn't working quite right.
>>> There is a shell command to call "pushd" (regress/Makefile.in), which
>>> doesn't exist in my user compiling PostGIS that uses bourne shell by
>>> default (not bourne again shell), so I didn't know if there were other
>>> things being called that don't exist on Solaris.
>>>
>>> I just ran make check and it failed 5 out of 42 tests, so I'll take a
>>> closer look at it.
>>>
>>> Thanks again for letting me know, and I'll definitely put that down as
>>> the thing to do.
>>> Nate
>>> _______________________________________________
>>> postgis-users mailing list
>>> postgis-users at postgis.refractions.net
>>> http://postgis.refractions.net/mailman/listinfo/postgis-users
>>>
>> _______________________________________________
>> postgis-users mailing list
>> postgis-users at postgis.refractions.net
>> http://postgis.refractions.net/mailman/listinfo/postgis-users
>>
> _______________________________________________
> postgis-users mailing list
> postgis-users at postgis.refractions.net
> http://postgis.refractions.net/mailman/listinfo/postgis-users
>