| Commit message (Collapse) | Author | Age | Files | Lines |
|---|
| | |
|
| | |
|
| |\
| |
| | |
Release prep
|
| | | |
|
| | | |
|
| |/ |
|
| |\
| |
| | |
Drop samefile and use a conditional dependency for old pythons
|
| |/
|
|
| |
Co-authored-by: Jason R. Coombs <jaraco@jaraco.com>
|
| |\
| |
| | |
Drop Travis
|
| |/
|
|
|
|
|
| |
* port check-readme to twine check before dist upload
* port self install tests to github
* artifact downloads
|
| |\
| |
| | |
Add release-branch-semver scheme
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
guess_next_simple_semver previously would error on trying to convert the
'rc0' part of a version such as '1.2.0rc1' to an int, even if it was
told to retain only the first two parts. (It would actually only error
on Python 2, where the conversion to int was non-lazy)
Now if it's told to retain only the first two parts, it won't try to
convert anything after the first two dots to an int, so it won't error.
The outcome is the same as if the bit after the second dot was a valid
int (which was going to be ignored anyway, we just now don't error on
it).
This is useful for the release-branch-semver scheme, which wants to be
able to find the next minor version even if the most recent version is a
release candidate.
|
| |/
|
|
|
|
|
|
|
| |
This is a scheme that assumes that the upcoming release for most
branches is a minor release, and that only for release branches
(those whose branch names parse as a version number according to current
configuration) is the upcoming version a patch release.
Better document the differences between the schemes in README.rst
|
| |\
| |
| | |
Skip git tests when git is unavailable
|
| |/ |
|
| |\
| |
| | |
Generate the date for today in UTC
|
| | | |
|
| |\ \
| |/
|/| |
Skip mercurial tests when hg is not available
|
| |/
|
|
|
|
|
| |
Mercurial is broken (read: crashing with bus error) on SPARC.
Nevertheless, many packages need setuptools_scm even if we never ever
use Mercurial checkouts. This patch makes it possible to run the test
suite successfully when Mercurial is not available on the system.
|
| | |
|
| | |
|
| |\
| |
| | |
closes #312 - document interaction with enterprise distros
|
| |/ |
|
| |\ |
|
| | |\
| | |
| | | |
Allow getting the version from the parent directory suffix.
|
| | | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | |
| | | |
Add a `parentdir_prefix_version` config parameter which allows lets
setuptools_scm get a version string from the name of the parent
directory: if `parentdir_prefix_version = "foo-"` and the parent
directory is named `foo-v12.34`, then the version is 12.34. This is
only active if the parameter is set (it defaults to unset, i.e. the
feature is opt-in), if getting the version from the SCM failed, *and*
there's no metadata already present. However, this has higher priority
than `fallback_version` (which remains the last-resort fallback).
This feature is intended to support the tarballs that GitHub
automatically creates for all releases, which have no metadata but
follow the correct naming scheme for the feature to work.
This feature is directly inspired from versioneer's `parentdir_prefix`
setting.
|
| |/ / |
|
| |\ \
| |/
|/| |
switch to github action
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
port testsuite to pathlib, and try to resolve windows short paths
trace git toplevel mismatch (observed on win32 name shortening)
exclude python2 on windows
remove the appveyor config, its broken
drop most travis envs that we have in github workflows now
add actions for creating egg artifacts/wheels/sdists
WIP: experiment with act
WIP: fix workflow typo
WIP: add missed runs-on
fix typoed excludes
fixup: upload actuall wheel/sdist in the build step
style fixes for the actions config
|
| |/ |
|
| |\
| |
| | |
Change ScmVersion.time to UTC
|
| | | |
|
| | | |
|
| | | |
|
| | | |
|
| |\ \
| |/
|/| |
Fix documentation for tag_regex and allow single-digit version numbers.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| |
| | |
The default `git describe` command had an undocumented `--match *.*` in it,
intended to skip over tags that may represent human-readable checkpoints.
Unfortunately this means it would be ignoring single-component versions.
Instead look for presence of digits in the tag, it's still a heuristic, but
it should cover all previous cases, and fix single component versions.
|
| | |
| |
| |
| |
| |
| |
| |
| |
| | |
This makes sure that even if there's only a single group of digits in the
version, it's correctly extracted, and any suffix ignored.
Also add tests for those particular cases to avoid regressions.
See Issue #411 for debugging details.
|
| | |
| |
| |
| |
| |
| |
| |
| | |
This appears to test something different than test_config.test_tag_regex,
but somehow also overrides the regular expression with a very old
(3-groups) one.
Remove the override and use the default regex instead (it works).
|
| |/
|
|
|
| |
Since 5853b0c72be9bf30701e58d9d72b413cb62ff1c9 (July 2018), the regex is
expected to have a single group, named `version` or unnamed.
|
| |\
| |
| | |
Drop the dot: py.test -> pytest
|
| |/ |
|
| |\
| |
| | |
no-local-version scheme and improved documentation
|
| |/ |
|
| | |
|
| |\
| |
| | |
#Fix 403: Close file descriptor before terminating process
|
| | | |
|
| |/ |
|
| |\
| |
| | |
Fix #399: ensure the git file finder closes file descriptor after reading archive
|
| | | |
|
