Subject: Re: RPATH (was Re: new packages)
To: None <email@example.com, firstname.lastname@example.org>
From: Ross Harvey <email@example.com>
Date: 08/10/1998 08:35:38
> From: Tim Rightnour <firstname.lastname@example.org>
> On 10-Aug-98 Hubert Feyrer spoke unto us all:
> # Couldn't we make some shorthand in our our toolchan for that, say - "-L"?
> Could we? Yes.
> Can we? Sure.
> Can we use it? Not unless you want to break all pre 1.4 builds on elf. It's
> *really* not that hard to do.. and the thing Todd just suggested would be an
> *excellent* addition to the pkg build rules. So then only non GNU-configure
> and non-Imake builds would need this fixing.
> Should we? I don't know.. This might be something to bring up with Ross.. (I'm
> CCing him on this) It certanly isnt feasable if we want to (read we *need* to)
> retain some backward compatibility with 1.3.
Well, -L specifically seems a bit unwise, as it should only be used for a
run-time rpath if no actual rpath directives were given, and it would be
quite confusing to specify such behavior when multiple directories were
specified. (Specified but not found? Specified and then found? And what
if we wanted to link with new build-tree libraries but search just the
defaults? Now you need a --no-auto-L. Hmm. Rather complex.)
If you only meant that as a quick example of a pretty-looking option, then
that brings up a Posix issue. I don't keep Posix specs around other than
1003.1 but IIRC -Wx,yyy,zzz... is a Posix-blessed syntax for pushing options
yyy,zzz... through to pass x, and now we are only talking about saving
two characters, right?