Message-ID: <87adcuzc3y.fsf@cain.internet2.edu>
Date: 07 Jun 2003 01:32:33 -0400
From: stanislav shalunov <shalunov@...ernet2.edu>
To: Solar Designer <solar@...nwall.com>
Cc: owl-users@...ts.openwall.com
Subject: Re: stmpclean problem
Solar Designer <solar@...nwall.com> writes:
> Currently, the chdir("/") is also done to ensure logging uses absolute
> pathnames. This means that you will need to do realpath(3) or an
> equivalent on any relative pathnames for logging, right?
Right. So I wanted to avoid that rathole by requiring absolute
pathnames on the command line (realpath() should work fine on Linux,
but, being a nonstandard---even if widely available function---might
not be there on other systems; besides the CAVEATS section of its man
page on FreeBSD mentions that Solaris realpath() `will, under certain
circumstances, return a relative resolved_path when given a relative
pathname').
The current behavior of silently treating relative paths as relative
to `/' clearly violates the Principle Of Least Astonishment. It
should not be violated by bailing on relative pathnames and should not
inconvenience users too much.
> [permissions=1777 check] may be a good idea, but it will break valid
> uses, including even the default use on Owl:
OK, you convinced me that it might not be a good idea after all.
Clearing man catfiles (and, I now notice, teTeX METAFONT-generated pk
fonts) requires stmpclean-like functionality, so it makes sense to
make sure stmpclean can do that job.
How about I check for `/' as the first char of the pathname from the
command line and refuse to run if it's not there?
--
Stanislav Shalunov http://www.internet2.edu/~shalunov/
This message is designed to be viewed upside down.
