bsd.sites.mk: Add GH_CHERRYPICK variable

[zirias]

Summary

If set to a list of commit hashes, dist patches are automatically
configured for fetching and applying from Github.

While here, add 'main' to disallowed values for GH_TAGNAME.

[zirias]

Rationale: it happened to me a few times that I had to patch upstream source for FreeBSD, and my patch was later accepted and merged. For github projects, this would add a simple way to apply the patch from upstream as long as it's still missing from the latest release.

[diizzy]

This seems to overcomplicating it?

See https://cgit.freebsd.org/ports/tree/audio/wavpack/Makefile for example

Edit:

There are also times where you might want to use another repo too such as

https://cgit.freebsd.org/ports/tree/irc/irssi/Makefile
https://cgit.freebsd.org/ports/tree/multimedia/dvdread/Makefile

[zirias]

Sure, but isn't "same repo" the common case? I'd argue it simplifies that case, you get rid of hand-crafted PATCH_SITES, PATCHFILES and PATCH_DIST_STRIP (or extra arg for the files)... and if GitHub ever changed how to access patches for individual commits, it could be handled in a central place.

[mat]

Ok, this was already discussed a few times, something called GH_PATCH I think.

The way to go is to put the patches in the files directory in the port, and git add them.

[lwhsu]

In D35782#812003, @mat wrote:

Ok, this was already discussed a few times, something called GH_PATCH I think.

The way to go is to put the patches in the files directory in the port, and git add them.

I can't find GH_PATCH in my mail archive or from google search. Would you like to share what has been discussed and the rationale of this conclusion? It is nice to be recorded and will help the future us.

[zirias]

In D35782#812003, @mat wrote:

The way to go is to put the patches in the files directory in the port, and git add them.

Is this common agreement? I often found the opinion that local patches should be the last resort. Pulling patches from GitHub seems to be common practice (and well, if it's a dist site and the ports framework offers dist patches, this makes sense to me).

[mat]

In D35782#812010, @lwhsu wrote:

In D35782#812003, @mat wrote:

Ok, this was already discussed a few times, something called GH_PATCH I think.

The way to go is to put the patches in the files directory in the port, and git add them.

I can't find GH_PATCH in my mail archive or from google search. Would you like to share what has been discussed and the rationale of this conclusion? It is nice to be recorded and will help the future us.

The rationale is building stuff is faster if the patch is already there, rather than having to fetch a few kilobytes of text.

[mat]

In D35782#812011, @felix_palmen-it.de wrote:

In D35782#812003, @mat wrote:

The way to go is to put the patches in the files directory in the port, and git add them.

Is this common agreement? I often found the opinion that local patches should be the last resort. Pulling patches from GitHub seems to be common practice (and well, if it's a dist site and the ports framework offers dist patches, this makes sense to me).

The problem is mainly depending more and more on external sources to get to the point where something can be built. Also, it is much faster to download a few patches as a git pack than having to run fetch for each of them.