PostGIS 3.5.0alpha2 released supports PG17

Greg Troxel gdt at lexort.com
Mon Jul 8 13:00:48 PDT 2024


"Regina Obe" <lr at pcorp.us> writes:

>>   make test has one failure, and I don't think this is a regression,
>>   just something I have not chased down.  But really this is pretty good.
>> 
>>       Test: test_spheroid_distance ...FAILED
>>       1. cu_geodetic.c:1355  -
>> CU_ASSERT_DOUBLE_EQUAL(d,10001965.7293127228117396,epsilon)
>
> Hmm has this one always failed, or this is the only one so better than past?
> I think we did change geodetic, Paul remember or we didn't except in a micro
> change, so micro releases might be impacted as well

I really don't remember clearly.  I can check.  My guess is that it is
not new and that it is probably some fiddly FP stuff not a real error.
I meant to say "don't worry about it; having 1 fail is great", on a
platform that the dev team is not chasing test fails on.

>>   NEWS says "DocBook5 XSL is now required to build html (Sandro
>>   Santilli)".  pkgsrc has docbook-xsl-1.79.2.  README.postgis gives a
>>   debian package name but doesn't describe it, and there's a "-ns"
>>   suffix.  So this is sort of a doc bug report.  Looking quickly it
>>   seems that the docbook-xsl situation is messy.   Is it really the case
>>   that almost every packaging system has this already?
>
> I think the thinking for this is now that we no longer require install of
> comments to build extensions building with docs is not absolutely necessary.
> And if people really want the docs, they can always pull them from news
> release links.
> Speaking of which I still have to figure out why Debbie is not building the
> comments.

Well there are multiple issues:

  README.postgis is unclear on what docbook5 xsl is, and given the
  confusion I think it should give a url to the project homepage.  Maybe
  I'm the only one unclear.

  If the build is configured to build the docs, it should fail if that
  fails.  I am ok with docs default off, and off if prereqs not found.

  It seems docbook5 xsl is needed but configure didn't fail what I have
  and decide not to build docs.  It tried to build and failed.

  docbook5 may or may not be an ok requirement.  I would say it's only
  ok if 95% of packaging systems already have it in what is considered
  their normal release.   Basically, a program should only depend  on
  things that either come with them (geos) or things that the world
  considers normal and not bleeding edge.


More information about the postgis-devel mailing list