| 1 |
On Wed, 11 Nov 2020 19:38:35 -0500 |
| 2 |
Rich Freeman <rich0@g.o> wrote: |
| 3 |
|
| 4 |
> I thought that the whole mirror:// thing was discouraged. For |
| 5 |
> anything else you just stick SRC_URI in the ebuild and the mirrors |
| 6 |
> should fetch it when they see it in the repo. |
| 7 |
> |
| 8 |
> I just host stuff like that on my dev webspace, or better yet on |
| 9 |
> github or something else that will auto-tarball stuff. |
| 10 |
|
| 11 |
In my last foray into building things from historical portage |
| 12 |
snapshots, the use of "push stuff directly to mirror:// and then only |
| 13 |
have mirror:// in SRC_URI" pattern was *the* leading cause of "oh noes, |
| 14 |
patches and distfiles no longer exist". |
| 15 |
|
| 16 |
Because naturally, once there is no longer an ebuild that requires the |
| 17 |
distfile, it can be reaped from the mirror, and this leaves no fallback |
| 18 |
fetch location. |
| 19 |
|
| 20 |
I found myself reviling this so much I feel it apt that this pattern |
| 21 |
should be more than discouraged, it should be banned outright and all |
| 22 |
ebuilds that *only* have mirror://gentoo/ as a source for any file |
| 23 |
should be shot repeatedly till it doesn't do that anymore. |
| 24 |
|
| 25 |
The resilience of Gentoo-authored distfiles should be *as good* as |
| 26 |
random 3rd parties we currently source from, not *exponentially worse* |
Archives