[tlbuild] very minor makefile fixes

[Karl Berry]

Hi Vincenzo - back on your mail from last June (sorry):

    https://tug.org/pipermail/tlbuild/2024q2/005555.html
    The attached patches fix the following bugs, which can be triggered with
    certain configure flags

I installed all those patches (r73540). Did not see any problems or make
any notable changes (just updated copyright years and made ChangeLog
entries).

@tlbuilders: although these changes should no effect on our "native"
builds, if you feel like giving the current source a tree a whirl, that
could only help avoid unpleasant surprises. I hope we will start trial
builds on Feb 1 and the pretest on Feb 8 (i.e., build the initial 2025
binaries in the first week of February).

   vm> Other issues for which I am happy to send tiny patches, but make
   vm> me nervous about side effects I did not consider:

   - etex should depend on $(proglib) -- once you do that, you can move
   lib.a, libp.a to EXTRA_LIBRARIES and stop recursing into lib (this
   marginally improves parallel building)
   - the main configure script checks for kpathsea before giving
   KPSE_ALL_SYSTEM_FLAGS a chance to fill the relevant variables with
   pkg-config (and the test should be made identical to the one in texk/web2c/
   configure.ac)

I'm happy to try both of these, since you're offering to send patches
(hope the offer did not expire :). I doubt they will cause problems for
our "native" builds. Especially the second, since as far as I know we
don't use pkg-config.

   vm> Then there are some weird things for which I have no suggestions,
   vm> all about web2c.

   - web2c/configure.ac: the logic around PWEB, UPWEB is off: the current
   AS_IF is always false so PWEB, UPWEB always coincide with $enable_webprogs

I think there's a bunch of logic errors in the enable/disable of various
parts of web2c. It's just never going to get to the top of my list since
we almost always build everything, or (for debugging) everything except
one thing.

   - web2c/configure.ac: WEB2C_DISABLE is called on 'luatex' and
   'luatex53', but it should be called on all lua[jit][hb]tex variants,
   I think?

Yes. And then the list of all the luatex variants should somehow be
parameterized instead of repeated a bunch of times for each missing fn.

BTW, I guess those fns exist on all modern systems, since no one has
ever reported "it tried to build luajittex even though I don't have
hstrerror", etc.

   - hitex does not cross compile nicely, because it compile and runs
   himktables

As Martin said, we can commit the .c file.

   - web2c/configure.ac has some counterintuitive test about
   $enable_omfonts and omegaware (you can trigger a build of otangle by
   *disabling* omfonts)

As above.

Fixes would be most welcome for any of the above, if you feel like
looking into any of them. Help with the configure setup is quite rare,
so I especially appreciate your efforts! --thanks, karl.