[Bug ada/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[Bug ada/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[pinskia at physics dot uc dot edu]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552


pinskia at physics dot uc dot edu changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Severity|critical                    |normal

[Bug ada/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[charlet at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552


charlet at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|ASSIGNED                    |WAITING


------- Additional Comments From charlet at gcc dot gnu dot org  2003-10-21 16:45 -------
Could you check again on the mainline, it may well be fixed after
the last Ada merge.

Arno

[Bug ada/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[ro at techfak dot uni-bielefeld dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552



------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2003-10-21 17:09 -------
Subject: Re:  make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

I had already tried this in various combinations, both with FSF GCC
mainline and various variants of the ACT GNAT CVS repository, and reported
the findings

	http://gcc.gnu.org/ml/gcc/2003-07/msg01388.html

without any reaction whatsoever.  So I fear this is still unchanged.

	Rainer

[Bug ada/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[ro at techfak dot uni-bielefeld dot de]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552



------- Additional Comments From ro at techfak dot uni-bielefeld dot de  2003-10-21 19:45 -------
Subject: Re:  make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

I had written my previous reply before I saw that the Ada merge had already
happened.  After a couple of fixes to get Ada to bootstrap at all (patch to
be submitted shortly), I still run into the original problem:

../../xgcc -B../../ -c -g      -W -Wall -gnatpg  a-chahan.adb -o a-chahan.o
+===========================GNAT BUG DETECTED==============================+
| 3.4 20031021 (experimental) (mips-sgi-irix6.5) GCC error:                |
| in handle_epilogue_set, at function.c:7525                               |
| Error detected at a-chahan.adb:434:14                                    |

	Rainer

[Bug middle-end/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[charlet at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552


charlet at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|bosch at gcc dot gnu dot org|unassigned at gcc dot gnu
                   |                            |dot org
             Status|WAITING                     |NEW
          Component|ada                         |middle-end


------- Additional Comments From charlet at gcc dot gnu dot org  2003-10-22 10:45 -------
Thanks for the follow up.
As indicated in the bug box, the crash is actually occurring in the
back-end so I am recategorizing this bug.

This is related to the handling of depressed stack pointer, so a work
around is to turn DSP off in 5gsystem.ads:

   Functions_Return_By_DSP   : constant Boolean := False;

Arno

[Bug middle-end/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[ro at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552



------- Additional Comments From ro at gcc dot gnu dot org  2003-10-24 20:41 -------
For some reason, this comment didn't make it into bugzilla via mail ;-(  So I'm
submitting it again via the web:

> This is related to the handling of depressed stack pointer, so a work
> around is to turn DSP off in 5gsystem.ads:
> 
>    Functions_Return_By_DSP   : constant Boolean := False;

With this workaround, I was able to bootstrap current mainline (C and Ada
only) on mips-sgi-irix6.5, though still with -O2 removed as described
below.  I'll use the freshly bootstrapped compiler to try another bootstrap
with -O2 and see how it goes.

Two problems: 

* gnatpsta fails to link (I think there's another PR for this bug):

ld32: ERROR   33 : Unresolved text symbol "mips_abi" -- 1st referenced by
../targtyps.o.
        Use linker option -v to see when and which objects, archives and dsos
are loaded.  
ld32: INFO    152: Output file removed because of error.
collect2: ld returned 2 exit status
gnatlink: cannot call ../../xgcc
make[2]: *** [../../gnatpsta] Error 4

* What is the consequence of this workaround?  From the comment in
  targparm.ads, it seems like setting it to False for now and returning to
  True later will break binary compatibility.

So how to proceed?  Is a proper fix likely to be in place for GCC 3.4?

	Rainer

[Bug middle-end/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[ro at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552



------- Additional Comments From ro at gcc dot gnu dot org  2003-10-24 21:06 -------
I have no idea what goes wrong with my mail submissions: it is getting really
tedious having to check if they make it to bugzilla and copy them to the web
interface otherwise ;-(

I've now been able to successfully run a full mips-sgi-irix6.5 bootstrap
with all languages including Ada with the workaround you mentioned.

The procedure was somewhat tedious, so here are the details:

* Bootstrap CVS mainline with GNAT 3.15p with -g removed from STAGE1_CFLAGS
  (to avoid PR ada/6669) and -O2 removed from the toplevel and gcc
  Makefiles to avoid a miscompilation by the bootstrap GNAT.

* Bootstrap CVS mainline again with the previously generated GNAT 3.4, but
  again with -O2 removed.

* Bootstrap CVS mainline a third time with full optimization.

This allowed both bootstrap and make gnatlib_and_tools to complete
successfully.

Besides, I tried the same thing on mips-sgi-irix6.2 to be able to provide a
bootstrap GNAT 3.4 on the oldest possible IRIX 6 version, but failed since
even without optimization, the generated stage1 compiler built by GNAT
3.15p ICEed at the end of stage1 building libgcc (sorry, I didn't keep the
details).

I may try again if I find a way to provide an IRIX 6.2 bootstrap GNAT,
otherwise I could make a provisory binary distribution of mainline GNAT for
IRIX 6.5 if you can tell me how to handle the Functions_Return_By_DSP
issue.

	Rainer

[Bug middle-end/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[ro at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552


ro at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |charlet at act-europe dot fr


------- Additional Comments From ro at gcc dot gnu dot org  2003-10-27 12:07 -------
Arnaud, would you please have a look at my last two comments?

Thanks.

    Rainer

[Bug middle-end/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[charlet at gcc dot gnu dot org]

PLEASE REPLY TO gcc-bugzilla@gcc.gnu.org ONLY, *NOT* gcc-bugs@gcc.gnu.org.

http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552



------- Additional Comments From charlet at gcc dot gnu dot org  2003-10-27 13:18 -------
Concerning gnatpsta: you just need to do a cvs update, since I've
disabled gnatpsta build in ada/Makefile.in recently.

Concerning DSP, this is really a middle-end regression that needs to be
investigated like anoy other middle-end bug. I am not familiar with
this code, so I can't comment further unfortunately.

The other possibly would be to commit the change to 5gsystem.ads
for now.

Arno

[Bug target/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[echristo at redhat dot com]

------- Additional Comments From echristo at redhat dot com  2004-12-01 20:40 -------
I doubt this is still a problem, but I'm going through bugs.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|NEW                         |WAITING


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552

[Bug target/6552] make gnatlib_and_tools fails compiling a-chahan.adb on IRIX 6.2

[echristo at redhat dot com]

------- Additional Comments From echristo at redhat dot com  2004-12-07 23:30 -------
Can reopen if still a problem.

-- 
           What    |Removed                     |Added
----------------------------------------------------------------------------
             Status|WAITING                     |RESOLVED
         Resolution|                            |FIXED


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=6552