I've tried this now myself and I can't reproduce the issue. It seems
poetry-core at least has no problem building with an older version of
itself installed, though this doesn't rule out some interaction with the
other ports being updated in the PR.
% port installed py310-poetry-core
The following ports are currently installed:
py310-poetry-core @1.0.8_0 (active)
% port info --version py310-poetry-core
version: 1.1.0
% sudo port -s upgrade py310-poetry-core
---> Computing dependencies for py310-poetry-core
---> Fetching distfiles for py310-poetry-core
---> Verifying checksums for py310-poetry-core
---> Extracting py310-poetry-core
---> Configuring py310-poetry-core
---> Building py310-poetry-core
---> Staging py310-poetry-core into destroot
---> Installing py310-poetry-core @1.1.0_0
---> Cleaning py310-poetry-core
---> Computing dependencies for py310-poetry-core
---> Deactivating py310-poetry-core @1.0.8_0
---> Cleaning py310-poetry-core
---> Activating py310-poetry-core @1.1.0_0
---> Cleaning py310-poetry-core
- Josh
On 2022-9-13 10:20 , David Gilman wrote:
So I note that if I drop the --no-isolation from the `python -m build
--wheel` that the python portgroup does I get a wheel built
successfully. I imagine the --no-isolation is there so normal packages
could take advantage of the MacPorts-provided build backend packages.
Maybe it is worth extending the python portgroup to accept some
Portfile option to omit the --no-isolation, which would only be used
on py-poetry-core and the other pep587 build backend packages. I would
hope that the build backend maintainers would maintain this sort of
cross-version compatibility but I could see this problem cropping up
under any of them. I think the ability to bootstrap its own wheel
compilation is not going to bitrot as I imagine they use it frequently
for their own CI and release builds.
On Mon, Sep 12, 2022 at 2:40 AM Joshua Root <j...@macports.org> wrote:
On 2022-9-12 06:22 , David Gilman wrote:
Log attached. It is a pure upgrade started with "port upgrade poetry".
The error in the log is actually the pep517 module complaining about the
backend:
:info:build Traceback (most recent call last):
:info:build File
"/opt/local/Library/Frameworks/Python.framework/Versions/3.10/lib/python3.10/site-packages/pep517/wrappers.py",
line 321, in _call_hook
:info:build raise BackendInvalid(
:info:build pep517.wrappers.BackendInvalid
:info:build ERROR Backend operation failed: BackendInvalid()
I don't know off the top of my head why this would happen. It might be
best to ask for help upstream.
- Josh
