On Wed, Oct 04, 2017 at 01:33:51PM +0200, Max wrote:
It is human choice - you'll have to explicitly
call "make release" to get the version
bumped.
What bumpversion does is helping to automate the boring details of it:
- determine current version (provided it matches the semver spec)
- pick next version number (according to semver spec)
How does it pick the next version number, i.e. how do I tell it whether to bump
major/minor/patch versions? Ah, I see in "Make a new release" wiki page:
make REL=minor release
Personally, I find it rather trivial to pick a number, which would also allow
manual version skips that we might see necessary for hypothetical reasons. As
in
make REL=2.0.1 release
If we did this, would we still use bumpversion for other tasks?
Semver validation could also be a simple regex.
If it's only us needing that program to make a release that's fine dependency
wise, so far I'm just probing what it is used for.
- commit the changes to git
It's not bumpversion that does the commit though, right. Anyway:
What changes get committed during 'make release':
- does it empty the TODO-RELEASE?
- does it bump LIBVERSION API versions? (probably not, right?)
- does it edit the debian changelog?
- tag the release commit
Ah, so the signed git tag is now done automatically? Also I see on the wiki
page that it may be necessary to re-tag the release after review, anyway. Maybe
tagging the release should not happen automatically? I'd rather not add my
official personal signature to a tag before reviewing. I also may need to pass
a GPG key ID to the tagging to sign with the proper key, I guess it's simpler
and less dangerous to leave it as a manual step...?
What do you think?
Thanks,
~N
P.S.: I'm still meaning to review the release hands-on and adjust the wiki
page, but I'm simply not getting all the things done that need attention at the
moment :/