| 1 |
On 17-09-2011 10:01:47 -0400, Rich Freeman wrote: |
| 2 |
> On Sat, Sep 17, 2011 at 5:07 AM, Nirbheek Chauhan <[1]nirbheek@g.o> |
| 3 |
> wrote: |
| 4 |
> |
| 5 |
> I agree. Plus, if we use git-notes to *append* to ChangeLog entries |
| 6 |
> (in case you left out important information), we've covered all bases |
| 7 |
> as far as I'm concerned. |
| 8 |
> |
| 9 |
> If something is REALLY critical that needs to be fixed you can always use a |
| 10 |
> hack like adding a line of whitespace or something and re-committing (which |
| 11 |
> will work with any scm), but I'd only reserve that for when a changelog |
| 12 |
> entry is so factually wrong that it risks a real problem. |
| 13 |
> |
| 14 |
> However, I'd really question whether ChangeLogs are the place to store such |
| 15 |
> a critical piece of documentation in the first place. If it is THAT |
| 16 |
> important that users know about a change it should be in news. |
| 17 |
> |
| 18 |
> I agree that it seems like we're at the point where "just a little better" |
| 19 |
> is becoming the enemy of "already way better than what we have." |
| 20 |
|
| 21 |
I would prefer going this route myself. Generate all ChangeLogs from |
| 22 |
commit messages only. This is easy to implement (POC is running for |
| 23 |
Prefix), but has a little issue with ChangeLog being in Manifest file. |
| 24 |
I think we should just omit it, or (better) allow the Manifest to have |
| 25 |
multiple signed parts, such that the ebuilds, dists and files are signed |
| 26 |
by the committing developer, and the ChangeLog is signed by the |
| 27 |
generation process (like snapshots are). |
| 28 |
|
| 29 |
The council has, however, (like Markos' initial mail explained) decided |
| 30 |
that ChangeLog entries must be updatable (e.g. corrected) afterwards. |
| 31 |
That makes a (ChangeLog) file IMO necessary, and hence nothing can be |
| 32 |
done different from making repoman update the ChangeLog itself upon |
| 33 |
commit one way or another. This isn't really new, it's bug #337853. |
| 34 |
|
| 35 |
Note that the latter approach also solves any problems of entry order, |
| 36 |
as are forseen when using a VCS like git. |
| 37 |
|
| 38 |
|
| 39 |
[1] https://bugs.gentoo.org/show_bug.cgi?id=337853 |
| 40 |
|
| 41 |
-- |
| 42 |
Fabian Groffen |
| 43 |
Gentoo on a different level |
Archives