This is the mail archive of the gcc@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: Question about egcs building w/native compiler


>>>>> nnorwitz  writes:

Neal> I'm trying to build the egcs compiler with the native cc (xlc) 
Neal> compiler on an IBM AIX 4.2 box.
     
Neal> I did a plain ./configure, then built with make bootstrap.
Neal> It gets pretty far (builds stage1, but can't build stage2), but 
Neal> eventually (in stage2) still tries to compile the fortran src.

Neal> I had another problem which I solved.  xlc needs an -ma flag to handle 
Neal> alloca() properly.  It would also work with a #pragma in the code, but 
Neal> that's a pain.  After adding the -ma flag it worked fine.  That should 
Neal> probably go in the configuration setup.
     
Neal> Is there an easy way to get around the fortran problem?
     
	I and others bootstrap EGCS on AIX 4.1, 4.2, and 4.3 without
difficulty, so there is some type of problem with your system and/or
cockpit error occurring.

	Your first line says bootstrapping with "native cc (xlc)".  Are
you really compiling with CC=cc or actually with CC=xlc?  The latter will
not work correctly as GCC/EGCS bootstrap relies on the native compiler
being invoked as "cc" if not GCC.  Your comments about requiring -ma for
alloca() also imply this same problem because GCC/EGCS supplies its own
alloca() implementation when using a system's native compiler and -ma
never is needed and never should be used for GCC builds.

	Only the GNU C compiler can be compiled with the native system
compiler when performing a bootstrap because special, heroic efforts have
been made to workaround native compiler deficiencies.  The rest of the
languages (C++, Fortran, Objective-C, etc.) all require GCC/EGCS to
boostrap.  I am confused by your comments about stage2 build failing yet
still trying to compile fortran src -- this cannot succeed.

	These problems seem to stem from the earlier failure and your
efforts to work around the problems under the assumption that the EGCS
bootstrap process is flawed on AIX 4.2.  The bootstrap process is correct
and works without problem; we need to figure out what is incorrect about
your environment that evokes the failure.

David



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