Re: Bug#189370: acknowledged by developer (irrelevant)
>> Atsuhito Kohda <email@example.com> writes:
> In this case, administrator should modify TEXINPUTS.latex in
> /etc/texmf/texmf.d/45TeXinputs.cnf and run update-texmf once.
> Generated texmf.cnf should be the same as an old one.
What makes you think that special casing the Debian installations is
something that every system administrator is willing to do?
I'm not saying that your solution is bad. It is fine as a generic
solution in the context of Debian, but it is not the only one.
> > Now image this: there's security upgrade for tetex-bin and the
> > poor fool has a cronjob that installs it. After its installation,
> > the local texmf.cnf is trashed.
> the local texmf.cnf is not trashed if administrator acted as above,
> only once.
That might be something the _system_ _administrator_ does not want to
do, for whatever reason, and that's none of _your_ bussiness. Repeat
this until you get it: "It's _his_ system, not mine."
> Am I missing something?
The forest for the trees, yes.
The point at discussion is simple: you can't blow away local changes
_without_ asking, and that's precisely what tetex-bin is doing. The
scope of the local changes is not your bussiness. Your only concern
should be "do I have permission to do this or not?" The default is
"no, I do not".
> BTW, in this case it is much better to do something;
> ln -s /random/path/texmf/tex/latex /usr/share/texmf/tex/latex/NFS
> then there is no need to modify texmf.cnf at all. (though I have no
> experience with NFS mounted directory)
You are not to question the system administrator's decisions.
(And JFTR, I tried to keep the example simple yet realistic to make the
point clear -- it seems I failed)
While we are at this, adding a /etc/default/texmf wouldn't be a bad
idea, IMO. There you could have something like
Marcelo | Item 13: List members in an initialization list in the
firstname.lastname@example.org | order which they are declared
| -- Scott Meyers, Effective C++