Patch to update libtool in GCC and Src trees

[Charles Wilson]

Paolo Bonzini wrote:
> As a stopgap solution, I would try adding an "_LT_DECL_SED" just above 
> the "if test "${use_libtool}" = yes; then" line.  This would be no big 
> deal compared to the LIB_AC_PROG_GCC hacks currently used by newlib.

True, if the $Xsed issue is the only one that crops up rather than 
simply the /first/ one.

Even so, you will still need to modify the Makefile.am's (if not all 75, 
then at least the 15 that correspond to the 15 configure.in's that 
invoke libtool).  I did all of them because (1) the extra -I doesn't 
hurt, and (2) making all newlib ACLOCAL_AMFLAGS follow the same pattern 
reduces complexity.

> However, I think the right way to do it is to mimic what the target 
> libraries in the gcc tree do: using AC_NO_EXECUTABLES and always using 

invoking the libtool initialization macros, but still avoiding actual 
use of libtool via USE_LIBTOOL?

or really, truly, always using libtool?

> libtool.  I can try to do so in the arm-elf configuration I use at work, 
> if it is okay to commit Steve's patch with the temporary measure above.

I'll give these suggestions a try tomorrow evening.

--
Chuck