Bug#176526: the problem with PATH setting in auxillary script of tetex
Thank you for the explanation-for-dummies
Walter Tautz <wtautz@cs.uwaterloo.ca> wrote:
By the way, could you have a look at http://bugs.debian.org/182880? We
think this has been fixed (but forgotten in the changelog
automatics). If you agree, please close it by sending any mail to
182880-done@bugs.debian.org
> cp: cannot create regular file `FAKE ACCESS COMMAND\n\n/var/cache/fonts/pk/ljfour/jknappen/ec/pk7825.tmp': No such file or directory
> [1] )
[...]
> I would suggest also looking at some of the other supplementary scripts in
> /var/lib/texmf/web2c/ if they also set their paths.... correctly
Okay, now I understand the problem and how it arises. However, I am not
convinced that it is a bug on our side. Instead of contemplating alone,
I share my doubts with you.
Within a given distribution, duplication of executable names is not
permitted. Thus, any program from this distribution can safely assume
that it will get what it wants if it calls "diff" or "sendmail". As a
side note, in maintainer scripts that are executed upon installation or
removal of a package, the use of binaries with pathnames is explicitly
discouraged.
So now you have something in your PATH that is called access and does
something different from tetex-bin's access. The situation is not so
fundamentally different from the problem you get if you have some
/usr/local/bin/sendmail that doesn't what /usr/sbin/sendmail does.
If this reasoning is right, than it is not a bug in tetex-bin's script
to not make sure they get _their_ access. It could even be regarded as a
bug if they did - assume that you have some weird filesystem where
access doesn't work, and you keep a patched one in /usr/local/bin...
This won't be very nice for you, but currently I feel this is how it
should be viewed.
What could be regarded as a bug, however, is the very generic name of
our binary. It could well be something like "fileaccess" or "tryaccess"
or "tex-access". On the other hand, what it does is, with some
decoration for option parsing, the most generic implementation of a
user-space binary using access(2).
Regards, Frank
--
Frank Küster, Biozentrum der Univ. Basel
Abt. Biophysikalische Chemie
Reply to: