1 |
On Mon, Mar 1, 2021 at 11:03 AM Neil Bothwick <neil@××××××××××.uk> wrote: |
2 |
> |
3 |
> On Mon, 1 Mar 2021 10:45:32 -0500, Jack wrote: |
4 |
> |
5 |
> > > Alternatively, switch to syncing from github and you'll always be as |
6 |
> > > up to date as possible - and it's much faster. |
7 |
> |
8 |
> > Syncing won't help until the ebuilds are fixed, and the bug does not |
9 |
> > say that has happened. The problem is not (yet) syncing your portage |
10 |
> > tree, it is with the mirrors, some of which have the old tarballs and |
11 |
> > some of which have the new tarballs. |
12 |
> |
13 |
> So it's the source mirrors that are out of date rather than the |
14 |
> rsync mirrors? |
15 |
> |
16 |
|
17 |
Not exactly. |
18 |
|
19 |
There are three files here: |
20 |
1. The ebuild/mainfest - which uses hash/size of an old version of the source. |
21 |
2. The distfile mirrors - which contain the corresponding old version |
22 |
of the source. |
23 |
3. The upstream SRC_URI - which contains a newer version of the source. |
24 |
|
25 |
Nothing has been fixed as of the time of sending this email, so you |
26 |
can sync your repo all you want and it will do no good. |
27 |
|
28 |
If you use a Gentoo distfile mirror you should be fine, because you're |
29 |
fetching an old version of the source, which matches what the ebuild |
30 |
is expecting. |
31 |
However, if you don't use a Gentoo distfile mirror then you're |
32 |
directly fetching the upstream source, which DOESN'T match the ebuild, |
33 |
and so you'll get an error. |
34 |
|
35 |
The Gentoo mirrors are going to reject any attempts to fetch the newer |
36 |
source, because they will only mirror a file if they match the Gentoo |
37 |
manifest. |
38 |
|
39 |
Usually these bugs do get fixed quickly, but there were changes in the |
40 |
upstream source files, so the maintainer probably wants to do some |
41 |
testing before updating the manifests. That is a GOOD thing and that |
42 |
is exactly why we have those manifests in the first place. They |
43 |
protect you, the user, in case upstream goes and changes a file out |
44 |
for any reason - Portage will reject this as this is not the file that |
45 |
was used by the package maintainer when doing all their QA activities. |
46 |
The biggest risk is malware getting snuck into an upstream file, but |
47 |
in this case it sounds like upstream changed something without |
48 |
versioning it. At the very least the maintainer is going to want to |
49 |
make sure it still works - you can't necessarily expect the same |
50 |
ebuild to work if upstream changes the sources. |
51 |
|
52 |
-- |
53 |
Rich |