GNAT 2.04 compilation problem under Linux

GNAT 2.04 compilation problem under Linux

[Sean McNeil]

Hello,
I was wondering if anyone can help me.  I get the following error
when I attempt to compile the rts lib for GNAT V 2.04:

+===========================GNAT BUG DETECTED==============================+
| Error detected at "s-arit64.adb", line 178:13                            |
| Please submit bug report by email to gnat-report@cs.nyu.edu              |
| Use a subject line meaningful to you and us to track the bug             |
| Include full sources in ASCII in a format compatible with gnatchop       |
| First line of sources must be marked by an Ada -- comment line           |
| Last line of sources must be last line of email message (no signature!)  |
| See gnatinfo.txt file for more info on procedure for submitting bugs     |
| GNAT V2.04 (i486-linux) Gigi abort, Code=999                             |
+==========================================================================+
compilation abandoned

gdb says:

(xxgdb) where
#0  0x1c8e93 in output_operand_lossage ()
#1  0x1c98fa in output_addr_const ()
#2  0x1f4e67 in print_operand ()
#3  0x1c950d in output_operand ()
#4  0x1c92ee in output_asm_insn ()
#5  0x1dc1ef in output_127 ()
#6  0x1c89e4 in final_scan_insn ()
#7  0x1c7dd9 in final ()
#8  0x122eb3 in rest_of_compilation ()
#9  0x7dc8 in end_subprog_body ()
#10 0xdc89 in tree_transform ()
#11 0xa999 in gnat_to_code ()
#12 0xf2e6 in process_decls ()
#13 0xe5c5 in tree_transform ()
#14 0xa999 in gnat_to_code ()
#15 0xa976 in gigi ()
#16 0x11eb03 in _ada_back_end ()
#17 0x11f179 in _ada_gnat1drv ()
#18 0x609b in yyparse ()
#19 0x121a5f in compile_file ()
#20 0x123f87 in main ()

Any ideas as to what might be wrong?

Sean McNeil

s-arit64.adb was Re: GNAT 2.04 compilation problem under Linux

[Tom Griest]

In article <smcneilD6nGz7.EpF@netcom.com> smcneil@netcom.com (Sean McNeil) writes:
>
>Hello,
>I was wondering if anyone can help me.  I get the following error
>when I attempt to compile the rts lib for GNAT V 2.04:
>
>+===========================GNAT BUG DETECTED==============================+
>| Error detected at "s-arit64.adb", line 178:13                            |

Sean,
  This is a recently discovered problem with the most recent release
of GNAT-2.04 on x86 architectures.  (It shows up on others besides
Linux.)  The folks at NYU are aware of the issue, and are working
on a solution.  Still, you should always CC: gnat-report@cs.nyu.edu
on a problem like this.

-Tom

Re: GNAT 2.04 compilation problem under Linux

[Robert Dewar]

Good grief, when you get a GNAT BUG DETECTED box with specific instrucions
to send an error report to gnat-report@cs.nyu.edu, it seems like it mighyt
be a good idea to follow it (if you already have sorry, I did not get a
copy of your mail yet).

In fact this was a very subtle problem that is i386 specific, and did not
show up on our OS/2 port to the i386. We finally rebuilt the DOS port,
replicated it, and are now testing out the patch, so stay tuned. 2.05
will of course have this patch, but also you will be able to get the
patch in the latest readme, available by the normal email method (se
GNAT documentation).

Re: s-arit64.adb was Re: GNAT 2.04 compilation problem under Linux

[Sean McNeil]

In article <3m3qn3INNpc3@ra.dept.cs.yale.edu>,
Tom Griest <griest-tom@cs.yale.edu> wrote:
>  This is a recently discovered problem with the most recent release
>of GNAT-2.04 on x86 architectures.  (It shows up on others besides
>Linux.)  The folks at NYU are aware of the issue, and are working
>on a solution.  Still, you should always CC: gnat-report@cs.nyu.edu
>on a problem like this.
>
>-Tom


Thanks Tom (and others) that have responded.  Just because I sent this
message to cla does not mean that I bypassed gnat-report@cs.nyu.edu.
I in fact did send a report, but made a shorter version for USENET
(not wanting to eat up bandwidth).  Another reason I posted was
because cs.nyu.edu was down last night for some reason (or my internet
feed via. netcom was down).

I just wanted to let everyone know that I CAN read directions :)

Sean McNeil

Re: GNAT 2.04 compilation problem under Linux

[Kenneth Almquist]

smcneil@netcom.com (Sean McNeil) writes:
> I was wondering if anyone can help me.  I get the following error
> when I attempt to compile the rts lib for GNAT V 2.04:

[ENTER Robert Dewar mode]

I'd like to point out that GNAT bugs should be reported to
gnat-report@cs.nyu.edu, since not all GNAT developers read c.l.a, and
those that do give higher priority to keeping up with gnat-report.  As
it happens, the bug you describe is a known bug which is currently
being investigated by the GNAT team.  It appears to be a bug in the
i386 code generator.  Until a fix is developed, you can use the following
work-around:

	When make starts compiling files in ada/rts, use ^Z to
	suspend make.  Change directory to ada/rts, and compile
	s-arit64.adb and s-tasclo.adb manually, using the GNAT 2.03
	compiler and the same options that make used to compile the
	other files in ada/rts.  Then unsuspend the make command.

[EXIT Robert Dewar mode]

It sure sounds like I know what I'm talking about, right?  :-)
				Kenneth Almquist

Re: GNAT 2.04 compilation problem under Linux

[Robert Dewar]

Kenneth's workaround is a neat kludge, and should work, however, we have
the patch, and are testing it at NYU here, should be out tomorrow.