Re: NLS files created in srcdir not in objdir

[Jeffrey A Law <law at cygnus dot com>]

  In message <20000330171048N.mitchell@codesourcery.com>you write:
  >     Philipp> Currently all NLS files generated, i.e. gcc.pot and the
  >     Philipp> .mo/.gmo files are created in the srcdir. This makes it
  >     Philipp> impossible to have srcdir mounted ro (e.g. via NFS) when
  >     Philipp> configured with --enable-nls. I believe this to be a bug
  >     Philipp> and if you agree, I'll try to come up with a patch that
  >     Philipp> creates those files in objdir.
  > 
  > I think that the policy should be that files that are the same, no
  > matter what target you're building, should go in the srcdir.  Other
  > things should go in the objdir.  I don't know which category the NLS
  > stuff falls into.  But, for example, the `parse.c' files are currently
  > put in the srcdir, and that prevents read-only mounting there, too.  I
  > think that's the right tradeoff -- it makes it easier to build more
  > than one target from the same source tree.
For CVS access, I'd really prefer to move away from having generated
files in the source tree (though exceptions may be reasonable).  In
particular I've actually had much better experiences with the parser
generated stuff as part of the build tree rather than the source tree.
[ For ftp snapshots/releases we would still include the generated
  files in the source tree. ]

I'm not sure where NLS fits in either, but I'd like to solve the parser
stuff before trying to tackle NLS since I've had more experience with the
parser side of the equation and it effects everyone if we make a change.

jef