[tex-live] texdoc can't find texmf.cnf

[Manuel Pegourie-Gonnard]

[re-sending on-list, sorry for the previous private reply, and adding a detail]

r8941 should fix this problem, though I'm not quite sure the root of the
problem is actualy fixed  (need to think a bit more about it).  Anyway, I
tested this on an OSX machine and it looks working.  Please complain if it
doesn't.

Jonathan Kew a écrit :
> Is it just me, or is there a problem with texdoc?

It's not you.  There's currently this problem on every non-linux unix
platform.  I'm working on it.

> Remember that the script is at
> texlive/Master/texmf/scripts/texlive/texdoc.tlu. So the variables here
> resolve to:
>
>   SELFAUTOLOC    -> texlive/Master/texmf/scripts/texlive
>   SELFAUTODIR    -> texlive/Master/texmf/scripts
>   SELFAUTOPARENT -> texlive/Master/texmf
>
They shouldn't.  They should resolve to the location of the symlink you
called, or the location of the actual texlua exacutable, which should be the
same (namely Master/bin/<platform>). Then the texmf.cnf can be found.

> AFAICT, there are two ways to address this: either (a) modify the
> compile-time path to something like [...]
> (untested, the intent was to make the entire /texmf{-local,} part
> optional); or (b) move the texdoc.tlu script, and any other scripts that
> might need to find texmf.cnf in a similar way, up one level in the
> hierarchy.
>
I don't think either of the two is that good, because it puts restrictions
on where the scripts can reside.  The best way is to have texdoc work as it
currently does on Linux, by making SELFAUTO* depend only on the location of
texlua.

> Or I suppose we could create a $SELFAUTOGRANDPARENT variable, and add
> this to the mix. I don't expect that to be a popular idea, though!
>
 :-)

Thanks for reporting,
Manuel.

PS: I'm glad to see people actually using texdoc  :-)