Stop using emacsmirror-mirror

[raxod502]

Per @tarsius in #356 (comment), the epkgs repository is now much smaller, and we have no need for mirroring it to https://github.com/emacs-straight/emacsmirror-mirror.

We can make sure the emacsmirror recipe repository in straight.el is still operational with the latest version of epkgs, and then switch the default value of straight-recipes-emacsmirror-use-mirror back to nil (& update associated documentation).

[raxod502]

(We should make sure the new repo is actually small enough to have negligible clone time; otherwise it'd make sense to continue with the current solution.)

[tarsius]

Lots of labels...

[raxod502]

Heh, I started tagging these issues a long time ago and I can't stop now! :D

[progfolio]

(We should make sure the new repo is actually small enough to have negligible clone time; otherwise it'd make sense to continue with the current solution.)

Mirror mirror on the wall, who's the fastest of them all?:

(ins)n@nbook:tmp $ time git clone git@github.com:emacs-straight/emacsmirror-mirror
Cloning into 'emacsmirror-mirror'...
remote: Enumerating objects: 535, done.
remote: Counting objects: 100% (344/344), done.
remote: Compressing objects: 100% (253/253), done.
remote: Total 535 (delta 142), reused 293 (delta 91), pack-reused 191
Receiving objects: 100% (535/535), 149.81 KiB | 2.67 MiB/s, done.
Resolving deltas: 100% (217/217), done.

real	0m0.627s
user	0m0.111s
sys	0m0.015s
(ins)n@nbook:tmp $ time git clone git@github.com:emacsmirror/epkgs.git
Cloning into 'epkgs'...
remote: Enumerating objects: 6067, done.
remote: Counting objects: 100% (1833/1833), done.
remote: Compressing objects: 100% (581/581), done.
remote: Total 6067 (delta 1407), reused 1673 (delta 1252), pack-reused 4234
Receiving objects: 100% (6067/6067), 22.85 MiB | 5.17 MiB/s, done.
Resolving deltas: 100% (4625/4625), done.

real	0m53.988s
user	1m25.971s
sys	0m1.482s
(ins)n@nbook:tmp $ 

I ran this several times for each and got similar results.
The mirror-mirror seems like it's still worth it to me.

[tarsius]

I git repack -a -d and pushed the result to a new temporary repository. Cloning that takes 25 seconds instead of 41 (instead of 2 for the mirror mirror).

This could get better once repacking kicks in on github's side but then it will gradually get worse again until...

So I think cloning this repository just takes about 1/2 to 1 minute and there's not much we can do about it. Whether that means you need to use a stripped down repository is up to you. IMO its okay if the initial clone is slow, subsequent fetches will be faster.

[raxod502]

I think given the infrastructure is already set up for the mirror-mirror, it makes sense to leave it in place unless it starts to create a maintenance burden.

[tarsius]

No harm in keeping that, but could you please update the documentation, which still claims:

straight-recipes-emacsmirror-use-mirror: Yes, there is also a mirror for Emacsmirror. This is because the epkgs repository contains a (frequently updated) SQLite database in it, which means the Git repository takes forever to clone (see #356).

It now contains a text file with sql statements that is used to re-generate the binary database locally.

[raxod502]

Done in 325f2ca.