> On 10 Jan 2014, at 12:14 pm, Adam Jensen <[email protected]> wrote:
> 
> > I think I empathize with your general sentiment but I'm not sure I
> > understand your proposal.
> > 
> > On Fri, 10 Jan 2014 09:58:05 +1300
> > David Koontz <[email protected]> wrote:
> >> 
> >> The idea of a release is to make a version of the code that
> >> doesn't
> >> change.  It represents a snapshot in development.  It implies
> >> after a
> >> release no one updates the released code.
> >> 
> > 
> > Yes! I would even extend that and recommend that there be
> > *additional*
> > snapshots, beyond just releases, that "freeze" significant moments
> > of
> > development and give them a meaningful label so people can refer to
> > these snapshots (and work with them if necessary).
> 
> Record the revision index!  Doing so is outside the scope of most
> revision systems.  The only way to freeze something (prevent someone
> one from writing to it) is generate a new revision (check something
> in).  Once there's a revision it's written in stone (without backing
> out changes.
> 
> This is from gna.org, the svn respository viewer.
> 
> I can checkout -r136 and get ghdl0.29.
> 
> With a branch like ghdl-0.31 there's literally nothing stopping a
> developer from adding to it, long after 'release'.  There is no
> branch locking mechanism.

Just to clarify again: this is not the purpose of a branch, but
the purpose of a tag.

> And notice 'Insert old NEWS file. in both the default branch and
> ghdl-0.31 (purple).

There is nothing wrong here: a bug can be fixed both in the release
branch and in the development branch.

> Also notice the revisions are sequence numbers for the repository and
> are unique across both branches.
> 
> Branches aside you're dealing with with one repository.
> 
> And contrary to the description revision 244 wasn't the last changes
> before 0.31 release.

To clarify again: 0.31 hasn't yet been released: no tag, no source
packages.

[...]

> The question arises how you keep someone other than the 'release'
> master from writing to the 'release' branch as opposed to the
> development branch.  If you have one person 'owning' the release
> they can back any changes someone else makes out as a matter of
> principle.
> 
> It would appear Brian (as the purported release master) let Tristan
> walk all over him.

My fault here. I should have and I will ask Brian before committing on
the release branch.

[...]

> 
> A run of the mill developer only needs to remember not to check in to
> the default branch during a release cycle. Whose responsibility it
> is to resolve the two after an actual release hasn't been defined.
>  The idea still needs a bit of refinement.
> 
> Don't volunteer to be a release master.  No need to main line
> caffeine.
> 
> Don't use a release candidate or release except from 'release'
> revisions.  No mumbling about when you checked out the branch for
> testing.
> 
> Naming a branch after a release is dangerous unless you have a way of
> preventing writes to that branch after a release.  This is what's
> being addressed here.

Should we simply use branch-0.31 (instead of ghdl-0.31) as branch name
to clarify that point ?

Tristan.

_______________________________________________
Ghdl-discuss mailing list
[email protected]
https://mail.gna.org/listinfo/ghdl-discuss

Reply via email to