Re: Trial release results on OpenServer

[Robert Lipe <robertl@dgii.com>]

Ian Lance Taylor wrote:
> I've put a trial release of binutils 2.9 at


I'll confess that I've been a bum about keeping binutils tested on
OpenServer.    Here are the last minute builds.   The "hard part"
of the code works, but it takes some fidgeting with the build.  This
isn't terribly unlike the way GAS has been on OpenServer for a few
years not but also isn't binutils' fault.   The big problem this time is
with libtool and I just haven't had time to fix libtool.   Since I also
haven't made GCC/EGCS talk to GAS (it requires the native assembler),
 this isn't as horrible as it could be.


CC=/bin/cc configure 
CC=/bin/cc configure --host-i586-pc-sco3.2v5.0.4coff --target=i586-pc-sco3.2v5.0.4coff

I overrode the build with /bin/cc instead of letting it pick up make 
becuase the libtooled code incorrectly passes '-belf' to CC regardless
of whether this is /bin/cc where -belf works or gcc where -melf would
be appropriate.

I had to edit '-lfl' out of the generated binutils/Makefiles.

(robertl) rjlhome:/play/tmp/elf
$  CC=gcc make -k check CC=gcc 

                === binutils Summary ===

# of expected passes            20
# of unexpected successes       1
# of untested testcases         3

FAIL: cofftag

                === gas Summary ===

# of expected passes            58
# of unexpected failures        1


The lone GAS failure was for 'cofftag.s'.   it doesn't seem unreasonable for
COFF input to choke an ELF assembler.




Then I built and tested it on COFF:

(robertl) rjlhome:/play/tmp/coff
$ CC="gcc -mcoff" make check CC="gcc -mcoff" 

                === binutils Summary ===

# of expected passes            20
# of expected failures          1
# of untested testcases         3

                === gas Summary ===

# of expected passes            59
../as-new 2.9


I dusted off the shell front-end script I did a few years ago to make the
two builds of GNU as emulate SCO's assembler well enough for EGCS/GCC to 
work  and sneaked it into my EGCS 1.0.1 gcc-lib area.   I built both
COFF and ELF "hello, world" programs successfully.   I never did get the
right combinations necessary to get a 'make bootstrap'  in an automated 
way to pass in the gas directories becuase of the -[mb]{elf,coff} thing.
I did, however, succeed in copying in the script file into stage/as
then doing a 
	make CC="gcc" CFLAGS="-O -Bstage/ -g " libdir=/usr/local/lib ALLOCA= as-new gasp-new

This worked for both COFF and ELF.   So I'm comfortable saying that it 
works as well on this target(s) as I'd expect it to.

Thanx to everyone involved; especially to Ian and his amazing ability to
make forward progress on this code and not break targets that he doesn't
use. :-)

-- 
Robert Lipe       http://www.dgii.com/people/robertl       robertl@dgii.com