compliation failure of minimal latex documents

[Phelype Oleinik]

Hi Jack,

On Tue, Feb 4, 2020 at 11:52 AM jack <tljack at forallx.net> wrote:
>
> I'm really hoping this is an oversight that's easily fixed in whatever
> part of core latex thinks it needs l3backend.  I don't see why a simple
> latex document with zero \usepackage calls needs to depend on experimental
> latex3 stuff.

The code layer of the LaTeX3 kernel (expl3) is no longer experimental
(even though the expl3 name suggests otherwise), and since the
2020-02-02 release, we pre-load expl3 in the LaTeX kernel (you can
read more about that in `texdoc ltnews31`).

However expl3 is loaded in the kernel *only if* expl3.ltx exists, which
means you have l3kernel installed, which belongs to
collection-latexrecommended, same as l3backend, so you should also
have l3backend, unless your installation is incomplete, for some reason.

> Moreover, the `tlmgr info` output for "l3backend" doesn't even tell me
> what collection (if any) it belongs to.  It should belong to
> collection-latex, though evidently it does not.

My `tlmgr info l3backend` ends with:
  collection:  collection-latexrecommended

> Recently there's been an alarming lack of attention paid to which
> collection things wind up in.  I'm thinking specifically of the recent
> oberdiek debacle: anything split from oberdiek should have gone in the
> same collection that oberdiek was in.  That's because some stuff in
> collection-latex depends on oberdiek.  But some of it wound up elsewhere,
> including collection-latexextra(!!!).  (That should really be fixed, BTW.)
>
> This meant that upon unexpected compliation failure of non-fancy documents
> I had to go hunt for these scattered oberdiek packages one-by-one,
> manually, instead of having them automatically updated along with
> collection-latex.  That is poor UX.

If I'm not mistaken you can send a list of these dependencies to
this mailing list and someone will take care of adding it to tlmgr...

Phelype