Re: gettext 0.10.37 msgfmt charset naming compatibility on Solaris 8

[Eli Zaretskii]

> From: Bruno Haible <address@hidden>
> Date: Thu, 17 May 2001 01:24:05 +0200 (CEST)
> 
> Paul Eggert writes:
> 
> > This is a bit discouraging, I'm afraid.  In effect, you're suggesting
> > that every GNU application run extensive tests on every vendor's
> > conversion tables when the application is built.
> 
> We also run tests whether vfork() works, and whether shared libraries
> work, when configuring GNU programs.

IMHO, the analogy here is wrong.  Testing for vfork on one machine
verifies that all other machines with the same installed revision of
the library will work or break in the same way.  So the program you
build is guaranteed to work correctly on other machines.

By contrast, testing for conversion tables which are not installed
yields an executable that cannot use the converters which were missing
at build time.  Since it is impractical to request that the build
machine has all of the converters with which the program can ever be
built, you are in effect saying that programs must be built on the
same machine on which they will be run, and that they need to be
rebuilt each time a new converter is installed, even if that converter
is simply some table that is loaded at run time, and even if the code
in the previously built executable already supports that table.

To me, that seems like a grave limitation.

Moreover, testing run-time behavior in a configure script generally
means that you cannot build a cross-compiled package.  It sounds like
testing the conversions at configure time is an even more fatal blow
to cross-building.