TL;DR:
Packaging workflow builds wheel incorrectly, listing incorrect dependencies.
pip install from PyPI will try to install wrong dependencies.
I have an issue installing trlc==3.0.0 under python 3.14.
We have discovered this when integrating your new release into the pyTRLCConverter and updating for 3.14 compatibility.
I can see from the CI workflow that 3.14 seems to run fine with trlc 3.0.0. This is because in the CI workflow, dependencies are installed by pip directly from the requirements-dev.txt, which includes the updated dependency to pyVCG==1.0.9.
The issue only appears when installing from the wheel published to PyPI.
When using a python 3.14 pip (or any pip version I guess) and using pip install trlc==3.0.0, pip will resolve to this: Collecting PyVCG==1.0.8 (from PyVCG[api]==1.0.8->trlc==3.0.0).
This is not a problem with previous python versions, but as pyVCG==1.0.8 is dependent on cvc5==1.3.0, which in turn requires python <3.14, the install will fail for python 3.14. This should be easy to reproduce by installing a python 3.14 and running pip install trlc==3.0.0.
While this is the symptom we experience when using python 3.14, the real issue seems to be in the packaging workflow.
When checking the METADATA for the trlc wheels published on PyPI, the dependency is indeed listed as pyVCG 1.0.8 (Requires-Dist: PyVCG[api]==1.0.8).
This is true for trlc 3.0.0, but I also checked for the 2.0.5 version, which had already switched to pyVCG==1.0.9. Here the wheel also lists a dependency for 1.0.8.
You can check this by downloading the wheel directly from PyPI and renaming the extension to .zip.
The included METADATA will list incorrect dependencies.
So somehow the workflow does not produce a package with the correct dependencies.
TL;DR:
Packaging workflow builds wheel incorrectly, listing incorrect dependencies.
pip install from PyPI will try to install wrong dependencies.
I have an issue installing trlc==3.0.0 under python 3.14.
We have discovered this when integrating your new release into the pyTRLCConverter and updating for 3.14 compatibility.
I can see from the CI workflow that 3.14 seems to run fine with trlc 3.0.0. This is because in the CI workflow, dependencies are installed by pip directly from the requirements-dev.txt, which includes the updated dependency to pyVCG==1.0.9.
The issue only appears when installing from the wheel published to PyPI.
When using a python 3.14 pip (or any pip version I guess) and using
pip install trlc==3.0.0, pip will resolve to this:Collecting PyVCG==1.0.8 (from PyVCG[api]==1.0.8->trlc==3.0.0).This is not a problem with previous python versions, but as pyVCG==1.0.8 is dependent on cvc5==1.3.0, which in turn requires python <3.14, the install will fail for python 3.14. This should be easy to reproduce by installing a python 3.14 and running
pip install trlc==3.0.0.While this is the symptom we experience when using python 3.14, the real issue seems to be in the packaging workflow.
When checking the METADATA for the trlc wheels published on PyPI, the dependency is indeed listed as pyVCG 1.0.8 (
Requires-Dist: PyVCG[api]==1.0.8).This is true for trlc 3.0.0, but I also checked for the 2.0.5 version, which had already switched to pyVCG==1.0.9. Here the wheel also lists a dependency for 1.0.8.
You can check this by downloading the wheel directly from PyPI and renaming the extension to .zip.
The included METADATA will list incorrect dependencies.
So somehow the workflow does not produce a package with the correct dependencies.