[gdal-dev] Motion: adopt GDAL 3.5.3RC1 as final 3.5.3 release

[Greg Troxel]

Greg Troxel <gdt at lexort.com> writes:

> I wasn't able to test this promptly, but am doing that now on netbsd-9
> amd64 with python 3.10.

And I am still using the autoconf build.

> gdal-lib (no python) built fine
>
> py-gdal (python bindings) is failing with what smells like a GNUism in
> sed patterns, but it's possible this is a change in pkgsrc not gdal and
> I haven't figured it out.

There are GNUisms in sed expressions to pacify coverity, and that is
skipped if it's not GNU sed -- but it's done at build time, not
configure time.   That causes a spurious:

  sed: unknown option -- -
  Usage:  sed [-aElnru] command [file ...]
          sed [-aElnru] [-e command] [-f command_file] [-I[extension]]
              [-i[extension]] [file ...]

which is just noise.   The next issue is:

  swig -Wall -I../include -I../include/python -I../include/python/docs -relativeimport -threads -outdir "osgeo"   -I/tmp/work/geography/py-gdal/work/gdal-3.5.3 -c++ -python -o extensions/gnm_wrap.cpp ../include/gnm.i
  sh: swig: not found

and this surprises me because I would expect swig to have been run
as part of 'make dist' or equiv, and I don't see configure looking for
swig.  Looking at
  https://trac.osgeo.org/gdal/wiki/BuildingOnUnix
I do not see swig listed as a prereq.


With the same system configuration, I reverted the change of version
from 3.5.2 to 3.5.3rc1, and then the py-gdal package built.


So I think something is wrong with 3.5.3rc1 and I would vote -1 on
release if I had a vote.

I would suggest that people test building the tarball on systems without
swig installed.  (with autoconf, as well as cmake)

-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 194 bytes
Desc: not available
URL: <http://lists.osgeo.org/pipermail/gdal-dev/attachments/20221027/b79b8e4b/attachment.sig>