Replace distutils throughout building system #351
+900
−1,212
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Closes #346
The main idea of this PR is to account for the deprecation and future removal of the Python package
distutils
. In doing so, I have encountered quite a few aspects that deserved maintenance, so I have tried to account for them as well. I describe all the changes below.Changes in
Makefile.in
and in thepython
folder reflect how building and installation of thefluidity
Python package are now handled throughsetuptools
,build
andpip
. Newpyproject.toml
andsetup.cfg
files replacesetup.py
in thepython
directory. These files describe the building process tobuild
, which then produces both a.tar.gz
archive and a.whl
wheel, stored in/python/dist
, that can be processed throughpip
, yielding a local installation of thefluidity
package. As a result, it is no longer required to updatePYTHONPATH
to have access tofluidity
,fluidity.diagnostics
,fluidity_tools
,GFD_basisChange_tools.py
andvtktools
, as reflected intools/testharness.py
.Additionally, in
Makefile.in
, I have replaced all instances ofpython3
with@PYTHON@
, which is sourced from the configuration stage (see below). Thepython_build
target is now properly handled and executed, andpython_clean
has been updated. I have also updated the output names of the testing scripts in line with changes to the CI (see below). Finally, theinstall
target does not executedefault
andfltools
targets anymore, which seemed quite repetitive. Similarly, thefltools
target now directly executes thefldecomp
target to avoid repetition.I have tried to simplify as much as I could the Python section of the
configure
file. I found that this section was quite confusing because it retained many compatibility aspects that are either deprecated or not supported anymore. The new Python section is located here. Additionally, I have removed the--enable-python
command-line argument that I believe no one uses.Regarding the CI, I have tried to simplify the Dockerfiles.
python3-venv
andbuild
are required for the package installation changes stated above. Other packages are handled through the virtual environment. I have removed the Groovy files, as Groovy is not supported anymore. For the.yaml
, I have pushed some changes in line with the 'testharnessClean' branch, which I am hoping to go back to when the more recent PRs are merged. These changes do not dramatically alter the way the CI operates, but I believe they bring some more clarity. Better handling of the longtests will be achieved through the 'testharnessClean' branch.Within
/libspud
, few directories use asetup.py
file. I have kept these files, albeit with calls tosetuptools
instead ofdistutils
, as not alldistutils
behaviours can currently be handled throughpyproject.toml
andsetup.cfg
files (see the note here). I have also accounted for changes in the NumPy C API within/libspud/python/libspud.c
.Finally, quite a few C++ scripts in
/tools
used the functionchdir
without having importedunistd
. I have corrected that.CI failures correspond to
spud-set
not being found at runtime (spud-set
is located at/libspud/bin
). I had already seen this behaviour when I was working on thetestharness
re-write, as highlighted here. I am puzzled at how the CI in themain
branch currently findsspud-set
in the first place.