[Debian-pan-maintainers] gemmi 0.7

[Andrius Merkys]

Hi Roland,

Thanks for reviving this discussion.

On 8/27/25 18:31, Roland Mas wrote:
> Le 27/01/2025 à 12:40, Andrius Merkys a écrit :
> Following up with that now that Trixie is out:
> 
>> On 2025-01-26 14:34, Roland Mas wrote:
>>> Could you list the reverse-dependencies in question so that they can 
>>> be fixed? I could add that gemmi <= 0.6.5 breaks at least one reverse 
>>> dependency 🙂
>>
>> The results of ratt rebuild for gemmi 0.7.0 from [1] are the following:
>>
>> * coot_1.1.09+dfsg-4 and dials_3.22.1+dfsg3-1: FAILED due to unsat- 
>> dependency: python3-numpy-abi9. These are due to an ongoing numpy 
>> transition.
> Building current coot with gemmi 0.7 works.

Good.

>> * finalcif_137+dfsg-2: FAILED due to incompatible function arguments 
>> when calling gemmi functions. There is a newer upstream release of 
>> finalcif, gemmi compatibility might be fixed there.
> Importing the new upstream release and updating a few details results in 
> a package that works and passes its testsuite. https://salsa.debian.org/ 
> lolando/finalcif has the code, seemingly ready to upload.

Nice, thanks. I will deal with finalcif as soon as gemmi 0.7.* becomes 
ready on buildds.

>> * freesasa_2.1.2-4: FAILED due to nonexisting gemmi function call. 
>> There is a newer upstream release for freesasa as well, but I recall I 
>> had other problems packaging it.
> As you mentioned, there hasn't been an upstream release recently. 
> However the code still evolves a bit (the last commit happened last 
> month); taking a snapshot from the master branch didn't immediately 
> result in a working package, even after two patches to rename structure 
> fields following Gemmi's renaming.

OK, I guess we can ignore freesasa for now and get back to it 
afterwards, hopefully with the help from upstream.

>> * macromoleculebuilder_4.0.0+dfsg-5: FAILED with "Gemmi version 0.7.0 
>> is not compatible with MMB. Major version must be exactly 0.6 and 
>> minor version .2 or newer."
> Still the case.

We may attempt patching the gemmi version check and see how far the 
build goes. But again, macromoleculebuilder has newer upstream release, 
which should be packaged post-trixie.

>> * molmodel_3.1.0-4.1: PASS
> 
> Still works.

Good.

> So in summary: coot, finalcif, molmodel and dials work with gemmi 0.7.0 
> (or can be made to work); freesasa and macromoleculebuilder don't appear 
> to.
> 
> Do you think this warrants uploading gemmi 0.7.0 to unstable, and maybe 
> get in touch with freesasa and macromoleculebuilder upstreams?

Thanks again for moving this forward. I have uploaded gemmi 0.7.1 to 
unstable just now. There is a newer release out now, but I did not dare 
to update to it yet. Let's first get the rdeps working for gemmi 0.7.1.

Best,
Andrius