This is the mail archive of the gdb@sourceware.cygnus.com mailing list for the GDB project. See the GDB home page for more information.

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

Re: gdb-19990209 snapshot fails to build with missing all target


   From: rodneybrown@pmsc.com
   Date: Mon, 15 Feb 99 14:05:23 -0500

   Snapshot gdb-19990209 fails to build because of missing opcode/Makefile
   all target, when configured on AIX-4.2.0, UnixWare 2.1.1 and SunOS 4.1.3.
   NB - while I'm patching up rather than pulling down a full tar.gz, the brik -Cb
   < gdb.crc
   worked & I did (belatedly) run the postpatch scripts.

I think this is all resolved now.  There is some experimentation going on
in opcodes-land...

   Further look at missing SOLIB_.* externals building on AIX-4.2.0
   I think that the relevant defines could be copied straight from the
   coff-solib.h and the include for xcoffsolib.h be pushed into a
   config/ header file.

Copying the definitions should work, because they are just placeholders
until somebody besides HP defines them usefully for their OS.  Actually,
looking at all of the *solib.h stuff, I seem some candidates for
simplification - one file (coff-solib.h) is never used, some are used
in inappropriate places, and they all go to a lot of trouble to use
macros, when a function pointer would be just as easy and not require
a bunch of #define/#include trickery.

   The reason I haven't patched this in is that
   gdb/eval.c includes xcofflib.h for a struct vmap definition.

This whole thing is moderately tangled - it's ugly for AIX because
the main program can be in a dynamically loaded library, which makes
trouble for GDB algorithms that expect to know where to find _start
and suchlike.  I hear rumors of IBMers hacking on all this anyway,
so there's not much point in going to be a lot of trouble on xcoff
solib stuff until we hear more about what they're up to.

							Stan