1 |
On Wed, Apr 15, 2015 at 12:27:08PM +0300, Andrew Savchenko wrote: |
2 |
> On Wed, 15 Apr 2015 11:06:22 +0200 Michał Górny wrote: |
3 |
> > Dnia 2015-04-15, o godz. 11:59:12 |
4 |
> > Andrew Savchenko <bircoph@g.o> napisał(a): |
5 |
> > |
6 |
> > > Hi, |
7 |
> > > |
8 |
> > > why manifest signatures are still optional for repoman? |
9 |
> > > |
10 |
> > > Repoman signatures are currently optional and this creates nasty |
11 |
> > > consequences: if signing errors occurs, repoman still proceeds :/ |
12 |
> > > |
13 |
> > > I just had a phone call during repoman commit and was not able to |
14 |
> > > type my password. Due to gpg-agent timeout repoman completed commit |
15 |
> > > without a signature :( Should signatures be mandatory, repoman will |
16 |
> > > bail out on such conditions and devs can recommit again safely. |
17 |
> > |
18 |
> > This is problem with the CVS two-commit procedure. The only solution is |
19 |
> > to stop using CVS keywords which people don't want to do because THEY |
20 |
> > ARE SO VERY USEFUL. |
21 |
> > |
22 |
> > Or make repoman do first commit without Manifest, so instead of |
23 |
> > unsigned Manifest you'd have Manifest failure. |
24 |
> |
25 |
> Is there any way to commit manifest first, then the rest of the |
26 |
> files? Of course there may be a network failure in-between, but |
27 |
> this should easily fixable with one more repoman commit run. |
28 |
|
29 |
no. the problem is that all the files change when they are committed. |
30 |
the $ Header: xxxxx $ line has the commit time and stuff. so the two |
31 |
phase first commits all the ebuilds and everything then when the lines |
32 |
are fixed it re-makes the manifest and commits. |
33 |
|
34 |
I dont think it can be fixed without getting rid of the $Header$ line. |
35 |
I'd be all for it, those lines seem like more trouble than its worth to me. |
36 |
|
37 |
-- Jason |