1 |
-----BEGIN PGP SIGNED MESSAGE----- |
2 |
Hash: SHA256 |
3 |
|
4 |
On 15/04/15 09:40 AM, Ian Stakenvicius wrote: |
5 |
> On 15/04/15 05:49 AM, Ulrich Mueller wrote: |
6 |
>>>>>>> On Wed, 15 Apr 2015, Micha³ Górny wrote: |
7 |
> |
8 |
>>> This is problem with the CVS two-commit procedure. The only |
9 |
>>> solution is to stop using CVS keywords which people don't want |
10 |
>>> to do because THEY ARE SO VERY USEFUL. |
11 |
> |
12 |
>>> Or make repoman do first commit without Manifest, so instead of |
13 |
>>> unsigned Manifest you'd have Manifest failure. |
14 |
> |
15 |
>> But that's what it does. It commits all other files, then it |
16 |
>> signs the Manifest and commits that. |
17 |
> |
18 |
>> However, if signing fails it will commit an unsigned Manifest. |
19 |
>> Which I think is a misfeature. If I have FEATURES=sign then I |
20 |
>> want to commit a signed Manifest. If there are problems, repoman |
21 |
>> should error out but not do some other action that I've not asked |
22 |
>> for. |
23 |
> |
24 |
>> Ulrich |
25 |
> |
26 |
> |
27 |
> Couldn't repoman sign a copy of the Manifest first (even if it's |
28 |
> staged in temporary space somewhere), then either do it all in one |
29 |
> commit or do the same two-stage commit it does now?? At least |
30 |
> that would allow it to catch gpg errors and abort. |
31 |
|
32 |
Nevermind, I follow the issue with the keywords now... |
33 |
|
34 |
-----BEGIN PGP SIGNATURE----- |
35 |
Version: GnuPG v2 |
36 |
|
37 |
iF4EAREIAAYFAlUuaykACgkQ2ugaI38ACPDWEQD/TX233xLS5CKkaikTSPEBaaNO |
38 |
ouaMvTubvt4LB+Vjdg0BALUgQD6be9hoj/c5IeFVED5X6WnLdlAbNnOLoUdn6kX0 |
39 |
=IreG |
40 |
-----END PGP SIGNATURE----- |