1 |
On 21:30 Sat 06 Nov , Donnie Berkholz wrote: |
2 |
> On 10:22 Sat 06 Nov , Krzysztof Pawlik wrote: |
3 |
> > I'm sending this patch for discussion, what it changes? The change is |
4 |
> > to where the final clone of repository will be placed, it used to be |
5 |
> > ${WORKDIR}/${module} (where module usually is the last component of |
6 |
> > source URI) to ${WORKDIR}/${P} (essentially ${S}). This has few |
7 |
> > effects: |
8 |
> > - ebuilds using mercurial.eclass don't need to set S any longer |
9 |
> > - mercurial.eclass behaves more like git.eclass |
10 |
> > - it breaks all existing ebuilds using this eclass |
11 |
> > |
12 |
> > The last effect is really, really nasty :( At the same time I think |
13 |
> > it's worth fixing this now, not when we have even more ebuilds using |
14 |
> > mercurial.eclass. Thankfully the fix to this is trivial: just remove |
15 |
> > the line where S is being set (or adjust it to match as is in case of |
16 |
> > one ebuild in the tree). |
17 |
> |
18 |
> Krzysztof, |
19 |
> |
20 |
> I see you've said you're breaking all these ebuilds but I don't see any |
21 |
> rationale for the change, at least in this email. Perhaps you could |
22 |
> explain why this level of breakage is necessary? |
23 |
|
24 |
I read it more closely and realized I was a little confused by the way |
25 |
you listed all the bullet points mixing together benefits and problems. |
26 |
|
27 |
So I'll try again: if you really want to do this change, you might want |
28 |
to consider adding a mercurial-2.eclass instead. Eclasses of this nature |
29 |
(svn, git, hg, etc) tend to be broadly used outside the tree as well as |
30 |
within, so breaking backwards compatibility can be a real problem. A new |
31 |
versioned eclass allows for a much more gradual transition. |
32 |
|
33 |
-- |
34 |
Thanks, |
35 |
Donnie |
36 |
|
37 |
Donnie Berkholz |
38 |
Sr. Developer, Gentoo Linux |
39 |
Blog: http://dberkholz.wordpress.com |