Re: GAS w/ egcs on OpenServer

[Robert Lipe <robertl@dgii.com>]

I've copied this back to the egcs mailing list becuase I'm getting
this question in my mailbox about three times a week now.  If I put
it on the egcs list, at least I can point to a URL in the archives.


Dag Nygren wrote:

> The 970929 snapshot seems to compile reasonably well
> using the  SCO as , but gas chokes on the -b coff /elf options
> and egcs seems to pass these even if I configure with --with-gnu-as.

The problem you are encountering was fixed yesterday by patches that 
I'll describe later.   Still, I get a lot of questions about using GAS 
on this target.

Currently gcc/egcs really wants to talk to the the SCO assembler.    It
isn't conscious of the --with-gnu-as flag.   The real problem is that 
GNU as can't handle changing between COFF and ELF at runtime becuase
the GAS COFF target isn't BFD-aware.

I have given some thought to adding support for the --with-gnu-as flag
and taking the weenie way out and supporting ONLY ELF with it.   (I'd 
also change the default configuration in GNU AS to emit ELF.)    It just
seems like too much bother to support the runtime switching of object
file formats.   Since the SCO assembler is available at no charge as 
part of the OSR5 Linkers and Libraries package, cross-dev is really the 
only compelling reason to use GAS anyway.   I'm not thoroughly convinced 
of the value of the work, but for the degenerate case, it shouldn't be 
a large effort.

> My question is: which as are you using, and is there a gas-version/frontend
> that groks the -b option ?

I've attached a documented that I submitted to the FSF for inclusion in 
GCC some time ago.   It wsa rejected and I just didn't care enough to
put up with the fight to get it accepted.

I should also confess that this document predates EGCS substantially.
The original version of it goes back almost a year.    It's entirely 
possible that we're now spitting out something that GAS doesn't like.
If so, we should look at it.



> PS. My make bootstrap stops when trying to compile the coff version
> of libgcc2 at the moment with the following??
>   ./xgcc -B./ -O2   -DIN_GCC    -g -I./include   -g1 -DIN_LIBGCC2 
> -D__GCC_FLOAT_NOT_NEEDED  -mcoff -I. -I. -I./config -c -DL${name} \
>       ./libgcc2.c -o ${name}.o; \
>   if [ $? -eq 0 ] ; then true; else exit 1; fi; \
>   ar rc tmplibgcc2.a ${name}.o; \
>   rm -f ${name}.o; \
> done
> _muldi3
> Assembler: libgcc2.c
>         aline 174       : Illegal subtraction
>         aline 191       : Illegal subtraction


You didn't apply the patches from yesterday in the thread titled:
	Patches for egcs/OSR5 Was: controlling exception type per target


Note that you also need to invert Jason't patch in that message.

In main(), the code inserted should be:

	#ifdef DWARF2_UNWIND_INFO
	exceptions_via_longjmp = ! DWARF2_UNWIND_INFO;
	#endif