| 1 |
On 12/07/2015 01:26 PM, Matt Turner wrote: |
| 2 |
>> |
| 3 |
>> I fixed it now. |
| 4 |
> |
| 5 |
> This is happening with some frequency. Could we install a server-side |
| 6 |
> git hook that prevents pushes if an ebuild is added without a Manifest |
| 7 |
> change? It wouldn't have to verify anything more than that to catch |
| 8 |
> nearly all of these problems. |
| 9 |
> |
| 10 |
|
| 11 |
The last time this happened I had the same idea and sat down to write |
| 12 |
one. The pre-push hook would do something like, |
| 13 |
|
| 14 |
* Get a list of cat/pkg affected by the push |
| 15 |
|
| 16 |
* Compare that list to profiles/categories to see which ones are |
| 17 |
packages (and should thus have manifests) |
| 18 |
|
| 19 |
* Verify the integrity of the files in each affected package by using |
| 20 |
something like `repoman manifest-check` |
| 21 |
|
| 22 |
I gave up when I realized that this final step requires network access. |
| 23 |
The missing manifest updates are for the distfiles, not the ebuilds. But |
| 24 |
you could probably still fake it. For example, if any ebuild files are |
| 25 |
added/removed, you expect Manifest to change. |
Archives