Conversation
|
If you agree with these changes, I would add them to release 1.4.2, what do you think ? |
Codecov Report
@@ Coverage Diff @@
## devel #348 +/- ##
=========================================
- Coverage 40.03% 39.7% -0.33%
=========================================
Files 79 81 +2
Lines 24128 26550 +2422
=========================================
+ Hits 9659 10543 +884
- Misses 14469 16007 +1538
Continue to review full report at Codecov.
|
|
is it ok with you if I try adding this to |
|
Since we have several PR to merge to master, I suggest we redo a release of 1.4.2 (delete the tag, retag, re-deploy) through a proper PR from devel to master. I would just like to solve issue #346 before that, hopefully this afternoon, ok for you ? Or, we can start now a cleaner workflow:
What do you think ? |
|
Sounds good to me :) |
This PR should fix a couple of issues:
Issue Anaconda deployment failing #340 : The version number in tofu/version.py had to be forced by hand
The problem came from the fact that travis checkouts the tag commit of tofu so the repository is not a specific branch but on a detached HEAD. I changed the
setup.pyso that it checks if it is building a travis tag version (by checking if$TRAVIS_TAGexists)tofu/setup.py
Lines 168 to 169 in d16595d
Issue Anaconda deployment failing #340 : Apparently the deployment was also done on test-pypi
As I mentioned in the issue, the condition
on: tags: Truesets TRAVIS_BRANCH to the tag number, thus it is no longermasternordeploy-test(nor any other branch name).Solved this by checking to which git branch the commit hash corresponds to:
tofu/.travis.yml
Lines 64 to 65 in d16595d
and then checking the result as a condition for deployment
tofu/.travis.yml
Line 96 in d16595d
There is one caveat for this to work properly, that should still be the proper way to release (will update the documentation):
It is necessary to first push the commit and then push the tag (if you only push the tag it will not find a branch corresponding to that tag so there will be no deployment). Thus, the workflow for a release should be something similar to:
Even if anaconda deployment was not being done, the recipe was still being build. which makes the whole process for nothing, so now it is being done in the upload script
Issue Unable to install from source on Windows #292 : the file
_updateversion.pywas no longer being deployed with pip:Unable to install from source on Windows #292 (comment)
I used the second solution: I took out out
build-backend = "setuptools.build_meta"inpyproject.toml