This is the mail archive of the gcc-patches@gcc.gnu.org mailing list for the GCC project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]

Re: Fix objc testsuite (was: Re: objc failures on branch)


At 11:55 08.03.2001, Andreas Jaeger wrote:
>Franz Sirl <Franz.Sirl-kernel@lauterbach.com> writes:
>
> > At 03:33 08.03.2001, Alexandre Oliva wrote:
> > >On Mar  7, 2001, Franz Sirl <Franz.Sirl-kernel@lauterbach.com> wrote:
> > >
> > > > Except that this currently won't work, as dejagnu only sets
> > > > LD_LIBRARY_PATH/SHLIB_PATH for running programs (unix.exp), but not
> > > > for compiling/linking (libgloss.exp) them :-(.
> > >
> > >Then how come the C++ testsuite works?
> >
> > Looking further in the sources...
> >
> > ah, because trickery somewhere passes "-lgcc_s -lc -lgcc_s" to
> > collect2/ld for g++, but for objc it's only "-lgcc -lc -lgcc". Somehow
> > the -lgcc ->
> >
> > -lgcc_s conversion doesn't kick in for objc (and other
> > languages?). Hmm, the c++ testsuite uses g++ as the driver, but objc
> > uses xgcc. I guess gccspec.c and/or need special casing for objc and
> > ENABLE_SHARED_LIBGCC, as it's done for g++ in cp/g++spec.c
>
>That conversion is done for C++ and Java (passing -shared-libgcc
>extra) but not for Fortran, C and Objective-C.  But how can it be done
>for C and Objective-C - and is this really the solution?
>
>It seems that the shared libgcc_s causes lots of problems for the
>whole GCC framework :-(

Yeah, seems so. Personally I like the idea of libeh, cause it restricts the 
shared stuff to the absolutely necessary. Maybe a -fPIC compiled libeh.a 
could be even pulled in 1:1 in during glibc build? Or the few sources 
involved can be shared between glibc and gcc, the same happens between 
other GNU packages too, or? Why wouldn't it work between glibc and GCC?

Franz.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]