From: Marc Espie <marc.espie.openbsd@gmail.com>
Subject: Re: redux Re: patch: update-plist + texlive
To: Edd Barrett <edd@theunixzoo.co.uk>
Cc: ports@openbsd.org, sthen@openbsd.org, tb@openbsd.org
Date: Tue, 7 Jul 2026 09:49:44 +0200

On Mon, Jul 06, 2026 at 08:38:54PM +0100, Edd Barrett wrote:
> -share/texmf-dist/scripts/minted/latex2pydata-0.5.0-py3-none-any.whl
> -share/texmf-dist/scripts/minted/latexminted-0.6.0-py3-none-any.whl
> +share/texmf-dist/scripts/minted/latex2pydata-0.5.0-${MODPY_PY_PREFIX}none-any.whl
> +share/texmf-dist/scripts/minted/latexminted-0.6.0-${MODPY_PY_PREFIX}none-any.whl
> ```

Yeah, MODPY_PY_PREFIX should probably be added to ignored variables.

> As you pointed out, we can now start culling some Python from FILESDIR.
> `write_plists.py` is now totally redundant and can be removed.
> 
> As I understand, this should also be possible:
> ```
> Index: files/update_plist_hints.py
> ===================================================================
> RCS file: /cvs/ports/print/texlive/texmf/files/update_plist_hints.py,v
> diff -u -p -r1.9 update_plist_hints.py
> --- files/update_plist_hints.py 8 Dec 2024 20:27:52 -0000       1.9
> +++ files/update_plist_hints.py 6 Jul 2026 19:15:43 -0000
> @@ -102,13 +102,6 @@ def build_subset_file_lists(tlpdb):
> 
>      sys.stderr.write("making plist map...\n")
> 
> -    # CONFLICTING PACKAGES
> -    # Whole packages that are ported elsewhere.
> -    conflict_pkgs = ["asymptote", "latexmk", "texworks", "t1utils",
> -                     "dvi2tty", "detex", "texinfo", "lcdftypetools"]
> -    conflict_pkg_files, conflict_symlinks = \
> -        collect_files(allspecs(conflict_pkgs, include_deps=False), db)
> -
>      # BUILDSET
>      # The smallest subset for building ports.
>      buildset_pkgs = [
> @@ -230,7 +223,7 @@ def build_subset_file_lists(tlpdb):
>          TargetPlist.DOCS: docs_files
>      }
> 
> -    comment_files = CONFLICT_FILES | conflict_pkg_files | commented_docs_files
> +    comment_files = CONFLICT_FILES | commented_docs_files
>      return plist_map, comment_files, symlink_map
> ```
> 
> (because that's what `UPDATE_PLIST_ARGS += -t ...` is all about)


Yep
>
> However, that change gives:
> ```
> Figuring out tie points
> Tieing loose objects
> Copying objects
> Not a HASH reference at /usr/ports/infrastructure/bin/update-plist line 1107.
> ```

Haven't seen that one ? 

tb's diff might miss a bit of mine

> (There's some more culling to do asides from that change, like `CONFLICT_FILES`
> can be trimmed down, maybe removed)
> 
> I'd be tempted to do:
> ```
> .PHONY: ${WRKDIR}/hints
> ```
Nope, it's not a .PHONY target.  This is bullshit.

> Because the target doesn't know when `update_plist_hints.py` has changed. I
> fell into this trap myself right off the bat. Thoughts?

Well, DUH, it's make.

Just have a 
${WRKDIR}/hints: ${FILESDIR}/update_plist_hints.py
dependency.

> Can we get a revised diff with the above perl crash fixed? Then I can do a
> proper job on the Python killing and run a partial bulk to look for fallout
> (although, I don't expect any tbh).
I will try my diff once I get the tb diff on my cvs mirror

> Later I will also look into why we are getting duplicate lines in the hints file. It
> wouldn't surprise me if there's a bug in the tlpdb causing that.

lol. It's not that important since I catch them