On Mon, Mar 06, 2023 at 11:36:49AM +0100, Christian Weisgerber wrote:
> Christian Weisgerber:
> 
> > I have a mirror mode got.git repository sitting here that I switched
> > from stsp_gitolite@git.gameoftrees.org to anonymous@got.gameoftrees.org
> > a few days ago:
> 
> I think that was a red herring.  I have another repository...
> 
> remote "origin" {
>         server stsp_gitolite@git.gameoftrees.org
>         protocol ssh
>         repository "/got-portable.git"
>         fetch_all_branches yes
> }
> 
> ... and it's showing the same thing.  I noticed that the last few
> fetches, although not slow, all downloaded suspiciously similar
> amounts of data:
> 
> -rw-r--r--  1 naddy  naddy   540647 Mar  3 22:52 pack-9e1a7580959dbe276a0c140fc4
> 7a9860866f813f.pack
> -rw-r--r--  1 naddy  naddy   530179 Mar  4 00:38 pack-1218978dbc2a61745df48c0086
> 1a6bd0e5059c6b.pack
> -rw-r--r--  1 naddy  naddy   505780 Mar  6 00:00 pack-80a1745fc7993cb8e5e9c280c1
> cfd2802f395be3.pack
> -rw-r--r--  1 naddy  naddy   530300 Mar  6 01:50 pack-9fb8e2f88076f36acf3fc4fd0b
> 16709ddf2de651.pack
> 
> Comparing "gotadmin ls | cut -d' ' -f1" of those pack files shows  
> a lot of repeated hashes.

That would suggest we have a bad bug in the client-side fetch code.
But with the reproducer you sent me I can reproduce this problem against
gotd, yet not against git-daemon/gitolite.

fetch main and 0.85 tag from gotd:
-rw-r--r--  1 stsp  stsp  -  160K Mar  8 07:33 objects/pack/pack-e0eb4f7a0245f0ee2d3e71cbd9786edda115ffd7.idx
-rw-r--r--  1 stsp  stsp  -  3.4M Mar  8 07:33 objects/pack/pack-e0eb4f7a0245f0ee2d3e71cbd9786edda115ffd7.pack

fetch main and 0.85 tag from git-daemon:
-rw-r--r--  1 stsp  stsp  -  5.7K Mar  8 07:36 objects/pack/pack-a8e14828f29e45b40c363ae55fe63ad901f3317b.idx
-rw-r--r--  1 stsp  stsp  -  302K Mar  8 07:36 objects/pack/pack-a8e14828f29e45b40c363ae55fe63ad901f3317b.pack