[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Bug#1023846: transition: gdal



On 11/23/22 19:22, Paul Gevers wrote:
On 23-11-2022 05:28, Sebastiaan Couwenberg wrote:
The libgdal-grass autopkgtest in testing is failing because it requires gdal, grass, and libgdal-grass from unstable.

  https://qa.debian.org/excuses.php?package=gdal

This combination needs to be tested to fix the regressions shown and unblock testing migration of gdal and the related rebuilt packages.

I'll schedule the set, but I have the feeling that a proper versioned relation (maybe an upper limit??) is missing somewhere. As there are quite a few versioned binaries involved already, it's obvious that there's a design. But if there's a runtime check for a version, ideally that should be expressed in dependencies too. Unless I'm missing something of course. (If that dependency would be there, britney would ask apt to pin packages from the source providing it from unstable if they are not fulfilled in testing).

"""
ERROR: Module built against version 2022-11-20T19:27:33+00:00 but trying to use version 2022-10-25T05:34:10+00:00. You need to rebuild GRASS GIS
or untangle multiple installations.
"""

This is not reflected in the dependencies, only the ABI dependency provided by grass-core is set:

 grass820

The dependency is set with a dh_gencontrol override.

This version check in grass is much stricter than it should be, the builds from the upstream git repo use the commit hash of include directory to check whether the code using grass is still compatible.

Because this requirement to rebuild libgdal-grass everytime grass is rebuilt annoyed me, I dug into this issue and had it changed upstream to use the full GRASS version (e.g. 8.2.0) like the virtual ABI dependency provided by grass-core for tarball builds.

GRASS 8.2.1 will contain this change, with their release process slower than expected, it's not sure whether it will be released before the bookworm freeze.

For future gdal transitions pulling in only the new gdal from unstable may suffice, although grass from testing still using the old gdal may cause different problems than just this version check. Like the crssync segfaults tend that happen with qgis when its dependencies are linked to different libproj versions.

"""
ERROR 1: OGR/GRASS driver was compiled against GDAL 3.5, but the current library version is 3.6 ERROR 1: GDAL/GRASS driver was compiled against GDAL 3.5, but the current library version is 3.6
"""

This is reflected in the libgdal-grass (1:1.0.2-2) dependencies:

 libgdal32 (>= 3.6.0)

Those are the normal ${shlibs:Depends} set via symbols files.

Kind Regards,

Bas

--
 GPG Key ID: 4096R/6750F10AE88D4AF1
Fingerprint: 8182 DE41 7056 408D 6146  50D1 6750 F10A E88D 4AF1


Reply to: