texinfo: 6.5 -> 6.7; fix cross compiling

[kampka]

Motivation for this change

The current derivation does not correctly cross-build when building in non-interactive mode.
Looking at the derivation, I don't see a reason for this distinction at all, hence this PR removes it.

Things done

• Tested using sandboxing (nix.useSandbox on NixOS, or option sandbox in nix.conf on non-NixOS linux)
• Built on platform(s)
  • NixOS
  • macOS
  • other Linux distributions
• Tested via one or more NixOS test(s) if existing and applicable for the change (look inside nixos/tests)
• Tested compilation of all pkgs that depend on this change using nix-shell -p nixpkgs-review --run "nixpkgs-review wip"
• Tested execution of all binary files (usually in ./result/bin/)
• Determined the impact on package closure size (by running nix path-info -S before and after)
• Ensured that relevant documentation is up to date
• Fits CONTRIBUTING.md.

[kampka]

@GrahamcOfBorg build texinfo

[kampka]

@thefloweringash according to git blame, you added the interactive cross-build check. Maybe you can shed some light on this?

[jonringer]

since this change causes over 500 rebuilds, do you mind targeting the staging branch

git rebase --onto=staging HEAD~1
git push .. ... --force

then change the base branch in the github PR from master -> staging

See https://nixos.org/nixpkgs/manual/#submitting-changes-staging-branch for more details on staging branch

$ nix-env -f /home/jon/.cache/nixpkgs-review/pr-95910/nixpkgs -qaP --xml --out-path --show-trace --meta
28975 packages updated:

[kampka]

@GrahamcOfBorg build texinfo

[kampka]

@jonringer Done.
Due to the massive rebuilds, I cannot claim that this change is well tested already, I can just claim that it builds on x86 and cross-aarch64 and that a couple of choice dependencies work as well (on master).

[oxij]

Non-interactive build mode is needed when using it for bootstrapping (otherwise it causes dependency loops, similarly to bash and stdenv) and `doCheckByDefault` nixpkgs feature. So do not remove it, please.

[Ericson2314]

if it is just used for bootstrapping, maybe the cross + non-interactive case won't occur very often in practice.

That cross patch looks very innocuous. Maybe we should try to always apply it? This is a mass rebuild already.

[kampka]

@oxij This change does not remove interactive mode, it just removes the flag that determines that cross-build patches are only applied on interactive mode. Now, they are applied on both, my question is, why where they separated in the first place in regards to cross building?

[thefloweringash]

@thefloweringash according to git blame, you added the interactive cross-build check. Maybe you can shed some light on this?

It was a while ago, but as far as I can tell only the interactive case was broken and it avoided a mass rebuild according to the tags on #76598. I don’t know what changed to make this now required for the non-interactive case, but it also fails for me on master.

Testing as of 6dd60c6

$ git checkout 6dd60c6cac8abc95f13f8bdd068d2c18865ca8b7
[...]
HEAD is now at 6dd60c6cac8 texinfoInteractive: fix cross build

$ nix-build -A 'pkgsCross.aarch64-multiplatform.texinfo' -A 'pkgsCross.aarch64-multiplatform.texinfoInteractive'
/nix/store/4pzd5442f0l43xzhwjfh7riir4pk872l-texinfo-6.5-aarch64-unknown-linux-gnu
/nix/store/d68f5rxc8582l2vz6b9c8rwkfjmybv9f-texinfo-interactive-6.5-aarch64-unknown-linux-gnu

[kampka]

Is there any way to get this patch improved to move it along? I know it's a bit last minute, but I'd really hate to miss the 20.09 release window with this ...

[FRidh]

This broke darwin stdenv and is blocking staging-next (and branch-off thereby) https://nix-cache.s3.amazonaws.com/log/lnkwicmp12bcy9fhdillhkwh2xm9px8b-texinfo-6.7.drv

[FRidh]

Proposed revert #97218.

[kampka]

Proposed revert #97218.

This PR contains two commits, the cross-compile fixes and the version bump, the later of which seems to introduce this regression.
I think it should be fairly easy to patch this forward, but if you want to revert, would it make sense to just revert the version bump (I'm mostly interested in the cross-compile fixes anyways).
Since I cannot test against Darwin, I'm afraid I cannot offer much active help in form of testing.

[FRidh]

A fix is preferred, and otherwise its a revert. I was under the impression it was the interactive change, but did not give it much thought.

I can't test it myself either, like most of us...

[vcunat]

Or postpone this whole PR after 20.09? I can't see why it needs to be in there as a last-minute change. (I can't test Darwin either.)