This is the mail archive of the
cygwin
mailing list for the Cygwin project.
RE: New package: makeself-2.1.5-2
- From: "Nellis, Kenneth" <Kenneth dot Nellis at acs-inc dot com>
- To: <cygwin at cygwin dot com>
- Date: Thu, 29 Apr 2010 09:17:49 -0500
- Subject: RE: New package: makeself-2.1.5-2
- References: <4BD87C2F.1040203@redhat.com>
> From: Eric Blake
> Sent: Wednesday, April 28, 2010 14:19
> To: cygwin@cygwin.com
> Subject: Re: New package: makeself-2.1.5-2
>
> ...
> Perhaps unspoken, but I prefer suffix-less executables. Then I don't
> have to care whether they are binary or interpreted scripts. Besides,
> having a suffix makes it harder to reimplement in a different language
> (for example, suppose someone decided to rewrite makeself in C, python,
> or perl, instead of sh). So following debian practice of stripping the
> .sh suffix as part of the packaging effort seems reasonable (and in the
> meantime, perhaps you may also want to report this upstream as a bug
> they might want to fix).
>
> --
> Eric Blake eblake@redhat.com +1-801-349-2682
> Libvirt virtualization library http://libvirt.org
But doesn't Debian's practice create other problems? If I want
to write a portable script that calls one of these scripts, I
have to call them differently whether I'm on a Debian system
or not. (Other workarounds exist, of course, e.g., creating
sym-links so either name will work.) And, if the upstream man
page correctly references the script with the suffix, when
Debian strips the script's suffix, does it also make the
corresponding change to the man page?
IMHO--but who cares?--the correct thing to do is leave the
suffix alone as the author intended, but to lobby for a change
in practice.
--Ken Nellis