On Fri, Oct 03, 2003 at 09:38:51AM +0200, Lars Gullik Bjønnes spake thusly:
 
> Angus Leeming <[EMAIL PROTECTED]> writes:
> 
> | Martin Vermeer wrote:
> >> Undoubtedly can be made to work... but it is 'politically' wise? If we
> >> go the way that Lars proposed and recommend people sticking to 2.95 to
> >> install and use stlport, than even the little extra hassle of
> >> writing/editing a configure_stlport script when a simple --use-stlport
> >> option to configure would be the 'least surprise' expectation, is too
> >> much. (And yes, I know this is for developers, not end users. But that
> >> includes, e.g., distribution packagers handling hundreds of packages
> >> and both unfamiliar with and uninterested in the LyX config/install
> >> system's idiosyncrasies.)
> >
> | Let's see how this all pans out first. I suspect Lars will give in and 
> | commit a debugstream that works for you too...
> 
> ...but now he has a LyX that works...
> 
> -- 
>       Lgb

My first question is: does this only concern developers, or end users
too? i.e., do we have any reason to produce binaries for end users on
an old compiler, or is it possible to produce such a binary on a
modern system?

If it is only a developer issue, then I would posit that STLport is
preferable to kludges. I can write up a cookbook recipe to get other
developers on 2.95 up in no time. Producing binaries for end users
shouldn't be too hard either: the RPM will have to have dependencies
on libpthread and STLport, which will be both installed into system
directories in the library search path. So the end user doesn't even
notice it except as a couple of extra dependencies. RPM's for STLport
appear to exist (and we could even link statically if must be to avoid
this issue altogether.) (BTW would STLport itself have a dependency on
libpthread?)

The only possible issue I see with this is having to support two
different STL/iostreams implementations with each their curious
curiosities. But surely that's still better than supporting an old and
broken such in addition to the modern gcc3 one.

- Martin

Attachment: pgp00000.pgp
Description: PGP signature

Reply via email to