[parisc-linux] Re: 64-bit long double support.

[parisc-linux] Re: 64-bit long double support.

[John David Anglin]

> Is there any opposition to checking in the fix for 3?
> Speak now or forever hold your peace.

Do it before I change my mind...

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: 64-bit long double support.

[Carlos O'Donell]

On 11/16/06, John David Anglin <dave@hiauly1.hia.nrc.ca> wrote:
> > Is there any opposition to checking in the fix for 3?
> > Speak now or forever hold your peace.
>
> Do it before I change my mind...

Trimming libc-alpha from the list.

Joseph Myers caught me on IRC at work and commented:
"... if you want 128-bit IEEE quad long double support, soft-fp should
work fine ..."

While I believe "work fine" is a relative term, it would require more
work than we have time for right now. I think there are more fun
projects to work on than this. For example I think gdb could use some
TLC for hppa.

Not only that but I have to help bootstrap distros using NPTL.
*waves to ubuntu and gentoo guys*

Don't let me stop you from working on doing a 128-bit IEEE quad long
double support. I will gladly do the "symbol transition" dance if you
get it implemented in gcc.

Cheers,
Carlos.
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: 64-bit long double support.

[Mike Frysinger]

[-- Attachment #1.1: Type: text/plain, Size: 439 bytes --]

On Thursday 16 November 2006 22:48, Carlos O'Donell wrote:
> Not only that but I have to help bootstrap distros using NPTL.
> *waves to ubuntu and gentoo guys*

i tried gcc-4.1.1/glibc-2.5 for hppa already and i thought it was broken ... 
plus, it makes it hard doing cross-compiles when my 64bit hosts (which i 
wh0re a lot because they're the fast ones) induce ICEs when building 
glibc/uclibc
http://gcc.gnu.org/PR29114
-mike

[-- Attachment #1.2: Type: application/pgp-signature, Size: 827 bytes --]

[-- Attachment #2: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: 64-bit long double support.

[Mike Frysinger]

[-- Attachment #1.1: Type: text/plain, Size: 565 bytes --]

On Thursday 16 November 2006 23:10, Mike Frysinger wrote:
> On Thursday 16 November 2006 22:48, Carlos O'Donell wrote:
> > Not only that but I have to help bootstrap distros using NPTL.
> > *waves to ubuntu and gentoo guys*
>
> i tried gcc-4.1.1/glibc-2.5 for hppa already and i thought it was broken

ok, just tried glibc-2.5 + hppa + linuxthreads and that is for sure broken 
(the defines for handling locking is all broken in the release)

what is the minimal binutils/gcc versions for TLS support ?  binutils-2.17 + 
gcc-4.1.1 == no hppa TLS
-mike

[-- Attachment #1.2: Type: application/pgp-signature, Size: 827 bytes --]

[-- Attachment #2: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: 64-bit long double support.

[Mike Frysinger]

[-- Attachment #1.1: Type: text/plain, Size: 1056 bytes --]

On Thursday 16 November 2006 23:16, Mike Frysinger wrote:
> ok, just tried glibc-2.5 + hppa + linuxthreads and that is for sure broken
> (the defines for handling locking is all broken in the release)

yeah glibc-2.5 fails as well:
... -o /var/tmp/portage/cross-hppa2.0-unknown-linux-gnu/glibc-2.5
/work/build-default-hppa2.0-unknown-linux-gnu-nptl/csu/crtn.o
In file included from ../nptl/descr.h:30,
                 from ../ports/sysdeps/hppa/nptl/tls.h:63,
                 from ../include/tls.h:6,
                 from <stdin>:2:
../ports/sysdeps/unix/sysv/linux/hppa/nptl/lowlevellock.h: In 
function '__lll_mutex_trylock':
../ports/sysdeps/unix/sysv/linux/hppa/nptl/lowlevellock.h:113: error: 
expected ':' or ')' before '-' token
...

i leave adding snap updates to GMsoft ... i try to stick with releases since i 
have to help so many arches :/

> what is the minimal binutils/gcc versions for TLS support ?  binutils-2.17
> + gcc-4.1.1 == no hppa TLS

upgraded to binutils-2.17.50.0.6 and i got TLS with gcc-4.1.1
-mike

[-- Attachment #1.2: Type: application/pgp-signature, Size: 827 bytes --]

[-- Attachment #2: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: 64-bit long double support.

[John David Anglin]

> http://gcc.gnu.org/PR29114

This has nothing to do with the discussion about long double support.

>>>From the PR, this only occurs in a cross from 64-bit host to 32-bit target.
There's no 64-bit parisc-linux host...

It would appear that the adddi3 expander in pa.md isn't used for some
reason in the initial code expansion as it would reject the insn.  You
are in the best position to determine where the insn is being emitted from.
It could be being emitted from the generic code or the PA backend.  Try
breaks on emit_move_sequence, emit_insn, emit_insn_after or emit_insn_before
and see if you can find where the insn is being emitted.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: 64-bit long double support.

[Mike Frysinger]

[-- Attachment #1.1: Type: text/plain, Size: 1096 bytes --]

On Friday 17 November 2006 00:35, John David Anglin wrote:
> > http://gcc.gnu.org/PR29114
>
> This has nothing to do with the discussion about long double support.

i didnt say it does ... i responding to the "testing NPTL in distros" comment

> From the PR, this only occurs in a cross from 64-bit host to 32-bit target.
> There's no 64-bit parisc-linux host...

i said that in my e-mail:
"plus, it makes it hard doing cross-compiles when my 64bit hosts"

> It would appear that the adddi3 expander in pa.md isn't used for some
> reason in the initial code expansion as it would reject the insn.  You
> are in the best position to determine where the insn is being emitted from.
> It could be being emitted from the generic code or the PA backend.  Try
> breaks on emit_move_sequence, emit_insn, emit_insn_after or
> emit_insn_before and see if you can find where the insn is being emitted.

erm, i've never debugged gcc before ... my experience involves producing 
reduced test cases and getting others to verify/fix gcc :)

but if it'll be of help to people ...
-mike

[-- Attachment #1.2: Type: application/pgp-signature, Size: 827 bytes --]

[-- Attachment #2: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: 64-bit long double support.

[John David Anglin]

> > It would appear that the adddi3 expander in pa.md isn't used for some
> > reason in the initial code expansion as it would reject the insn.  You
> > are in the best position to determine where the insn is being emitted fro=
> m.
> > It could be being emitted from the generic code or the PA backend.  Try
> > breaks on emit_move_sequence, emit_insn, emit_insn_after or
> > emit_insn_before and see if you can find where the insn is being emitted.
> 
> erm, i've never debugged gcc before ... my experience involves producing=20
> reduced test cases and getting others to verify/fix gcc :)

If you add -v to the gcc command line, the driver will output the
command it uses to run cc1.  This is needed to debug cc1.

GCC is a volunteer based project, so help is always welcome.  I got
involved when similar issues affected the projects that I was working
on.

There's always bugs (e.g., mainline is currently broken for both 32
and 64 bits on hpux).  There are over 50 open PRs for hppa.  There
are more critical issues that need to be resolved for the upcoming
4.2.0 release.  Most people do native builds and the problem doesn't
appear there.  Fixing this requires finding a suitable 64-bit host to
debug it on.

This problem should be relatively easy to fix if we know where and
why the insn is being emitted.

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] Re: 64-bit long double support.

[Jeff Bailey]

[-- Attachment #1.1: Type: text/plain, Size: 1284 bytes --]

On 16/11/06, Mike Frysinger <vapier@gentoo.org> wrote:
>
> On Thursday 16 November 2006 23:16, Mike Frysinger wrote:
> > ok, just tried glibc-2.5 + hppa + linuxthreads and that is for sure
> broken
> > (the defines for handling locking is all broken in the release)
>
> yeah glibc-2.5 fails as well:
> ... -o /var/tmp/portage/cross-hppa2.0-unknown-linux-gnu/glibc-2.5
> /work/build-default-hppa2.0-unknown-linux-gnu-nptl/csu/crtn.o
> In file included from ../nptl/descr.h:30,
>                  from ../ports/sysdeps/hppa/nptl/tls.h:63,
>                  from ../include/tls.h:6,
>                  from <stdin>:2:
> ../ports/sysdeps/unix/sysv/linux/hppa/nptl/lowlevellock.h: In
> function '__lll_mutex_trylock':
> ../ports/sysdeps/unix/sysv/linux/hppa/nptl/lowlevellock.h:113: error:
> expected ':' or ')' before '-' token
> ...
>
> i leave adding snap updates to GMsoft ... i try to stick with releases
> since i
> have to help so many arches :/
>
> > what is the minimal binutils/gcc versions for TLS support ?
> binutils-2.17
> > + gcc-4.1.1 == no hppa TLS
>
> upgraded to binutils-2.17.50.0.6 and i got TLS with gcc-4.1.1
> -mike



If you want a reference implementation, we have this running in Ubuntu
Feisty now.




-- 
Jeff Bailey - http://www.raspberryginger.com/jbailey/

[-- Attachment #1.2: Type: text/html, Size: 1939 bytes --]

[-- Attachment #2: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] gcc ICE PR29114 (was: 64-bit long double support.)

[John David Anglin]

> i've attached my gdb run ... i'm not familiar with gcc so i dont know what 
> more information you need to see to debug

Excellent progress.

> Breakpoint 1, emit_move_sequence (operands=0x7c3908288900, mode=DImode, scratch_reg=0x0)
>     at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/config/pa/pa.c:1309
> 1309	{

What we need to see is the backtrace when emit_move_sequence (or emit_insn)
is called with operands that match those in the insn that causes the ICE below.
This will localize the area of the problem.

In the call to emit_move_sequence, operands is an array.  operands[0] is
the DEST for the move operation.  operands[1] is the SRC.

You can see the RTL for the first two operands with the commands:

p debug_rtl (operands[0])
p debug_rtl (operands[1])

In the break at emit_insn,

p debug_rtl (x)

will print "x" in formated form.

Other useful GCC debug commands:

p debug_rtl_list (INSN, N)
where INSN is an instruction RTX.  This will print N insns.

p debug_tree (TREE)
where TREE is a tree variable will print out the tree struct.

p operands[0]
will print operands[0] as a C struct.

ptype operands[0]
will show the RTX type details.  This is useful for figuring
out how to set a conditional break that will check for "reg:DI 95".

> ../seed48_r.i: In function 'main':
> ../seed48_r.i:1: error: unrecognizable insn:
> (insn 9 8 10 1 (set (reg:DI 95)
>         (plus:DI (reg:DI 96)
>             (const_int -6547 [0xffffffffffffe66d]))) -1 (nil)
>     (expr_list:REG_EQUAL (const_int 25214903917 [0x5deece66d])
>         (nil)))
> ../seed48_r.i:1: internal compiler error: in extract_insn, at recog.c:2084
> Please submit a full bug report,
> with preprocessed source if appropriate.
> See <URL:http://gcc.gnu.org/bugs.html> for instructions.
> 
> Program exited with code 01.
> (gdb) 

Dave
-- 
J. David Anglin                                  dave.anglin@nrc-cnrc.gc.ca
National Research Council of Canada              (613) 990-0752 (FAX: 952-6602)
_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] gcc ICE PR29114 (was: 64-bit long double support.)

[Mike Frysinger]

[-- Attachment #1.1: Type: text/plain, Size: 343 bytes --]

On Friday 17 November 2006 15:51, John David Anglin wrote:
> What we need to see is the backtrace when emit_move_sequence (or emit_insn)
> is called with operands that match those in the insn that causes the ICE
> below. This will localize the area of the problem.

updated trace with automatic pretty prints attached to each breakpoint
-mike

[-- Attachment #1.2: Type: application/pgp-signature, Size: 827 bytes --]

[-- Attachment #2: gdb-run --]
[-- Type: text/plain, Size: 20465 bytes --]

$ gdb
Using host libthread_db library "/lib/libthread_db.so.1".
Breakpoint 1 at 0x72eeea: file /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/config/pa/pa.c, line 1310.
Breakpoint 2 at 0x5878c5: file /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c, line 4433.
Breakpoint 3 at 0x587388: file /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c, line 4260.
Breakpoint 4 at 0x587670: file /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c, line 4354.
GNU C version 4.1.2 20061117 (prerelease) (hppa2.0-unknown-linux-gnu)
	compiled by GNU C version 4.1.1 (Gentoo 4.1.1-r2).
GGC heuristics: --param ggc-min-expand=100 --param ggc-min-heapsize=131072
Compiler executable checksum: 75c21c80b69b3dedab682219e007845e

Breakpoint 2, emit_insn (x=0x0) at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
(gdb) c
Continuing.

Breakpoint 1, emit_move_sequence (operands=0x70c27856bec0, mode=DImode, scratch_reg=0x0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/config/pa/pa.c:1310
1310	  register rtx operand0 = operands[0];
$1 = {code = MEM, mode = DImode, jump = 0, call = 1, unchanging = 0, volatil = 0, in_struct = 0, used = 0, 
  frame_related = 0, return_val = 1, u = {fld = {{rt_int = -1009418144, rt_uint = 3285549152, 
        rt_str = 0x3204c3d58060 "\"", rt_rtx = 0x3204c3d58060, rt_rtvec = 0x3204c3d58060, 
        rt_type = 3285549152, rt_addr_diff_vec_flags = {min_align = 96, base_after_vec = 0, 
          min_after_vec = 0, max_after_vec = 0, min_after_base = 0, max_after_base = 0, 
          offset_unsigned = 0, scale = 213}, rt_cselib = 0x3204c3d58060, rt_bit = 0x3204c3d58060, 
        rt_tree = 0x3204c3d58060, rt_bb = 0x3204c3d58060, rt_mem = 0x3204c3d58060, 
        rt_reg = 0x3204c3d58060}}, hwint = {54996046807136}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102673411, sig = {54996047739472, 0, 458786}}}}
$2 = {code = CONST_INT, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -554899859, rt_uint = 3740067437, 
        rt_str = 0x5deece66d <Address 0x5deece66d out of bounds>, rt_rtx = 0x5deece66d, 
        rt_rtvec = 0x5deece66d, rt_type = 3740067437, rt_addr_diff_vec_flags = {min_align = 109, 
          base_after_vec = 0, min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 0, 
          offset_unsigned = 1, scale = 236}, rt_cselib = 0x5deece66d, rt_bit = 0x5deece66d, 
        rt_tree = 0x5deece66d, rt_bb = 0x5deece66d, rt_mem = 0x5deece66d, rt_reg = 0x5deece66d}}, hwint = {
      25214903917}, rv = {cl = 1, sign = 1, signalling = 1, canonical = 0, uexp = 116877107, sig = {0, 0, 
        0}}}}
(gdb) 
Continuing.

Breakpoint 1, emit_move_sequence (operands=0x70c27856bb80, mode=DImode, scratch_reg=0x0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/config/pa/pa.c:1310
1310	  register rtx operand0 = operands[0];
$3 = {code = REG, mode = DImode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, used = 0, 
  frame_related = 0, return_val = 0, u = {fld = {{rt_int = 95, rt_uint = 95, 
        rt_str = 0x5f <Address 0x5f out of bounds>, rt_rtx = 0x5f, rt_rtvec = 0x5f, rt_type = 95, 
        rt_addr_diff_vec_flags = {min_align = 95, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0x5f, 
        rt_bit = 0x5f, rt_tree = 0x5f, rt_bb = 0x5f, rt_mem = 0x5f, rt_reg = 0x5f}}, hwint = {95}, rv = {
      cl = 3, sign = 1, signalling = 1, canonical = 1, uexp = 2, sig = {95, 0, 0}}}}
$4 = {code = CONST_INT, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -554899859, rt_uint = 3740067437, 
        rt_str = 0x5deece66d <Address 0x5deece66d out of bounds>, rt_rtx = 0x5deece66d, 
        rt_rtvec = 0x5deece66d, rt_type = 3740067437, rt_addr_diff_vec_flags = {min_align = 109, 
          base_after_vec = 0, min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 0, 
          offset_unsigned = 1, scale = 236}, rt_cselib = 0x5deece66d, rt_bit = 0x5deece66d, 
        rt_tree = 0x5deece66d, rt_bb = 0x5deece66d, rt_mem = 0x5deece66d, rt_reg = 0x5deece66d}}, hwint = {
      25214903917}, rv = {cl = 1, sign = 1, signalling = 1, canonical = 0, uexp = 116877107, sig = {0, 0, 
        0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3e31740)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$5 = {code = SET, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -1008527584, rt_uint = 3286439712, 
        rt_str = 0x3204c3e31720 "\"", rt_rtx = 0x3204c3e31720, rt_rtvec = 0x3204c3e31720, 
        rt_type = 3286439712, rt_addr_diff_vec_flags = {min_align = 32, base_after_vec = 1, 
          min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 1, 
          offset_unsigned = 0, scale = 227}, rt_cselib = 0x3204c3e31720, rt_bit = 0x3204c3e31720, 
        rt_tree = 0x3204c3e31720, rt_bb = 0x3204c3e31720, rt_mem = 0x3204c3e31720, 
        rt_reg = 0x3204c3e31720}}, hwint = {54996047697696}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102701241, sig = {54996046949248, 0, 0}}}}
(gdb) 
Continuing.

Breakpoint 1, emit_move_sequence (operands=0x70c27856b8c0, mode=DImode, scratch_reg=0x0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/config/pa/pa.c:1310
1310	  register rtx operand0 = operands[0];
$6 = {code = REG, mode = DImode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, used = 0, 
  frame_related = 0, return_val = 0, u = {fld = {{rt_int = 95, rt_uint = 95, 
        rt_str = 0x5f <Address 0x5f out of bounds>, rt_rtx = 0x5f, rt_rtvec = 0x5f, rt_type = 95, 
        rt_addr_diff_vec_flags = {min_align = 95, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0x5f, 
        rt_bit = 0x5f, rt_tree = 0x5f, rt_bb = 0x5f, rt_mem = 0x5f, rt_reg = 0x5f}}, hwint = {95}, rv = {
      cl = 3, sign = 1, signalling = 1, canonical = 1, uexp = 2, sig = {95, 0, 0}}}}
$7 = {code = PLUS, mode = DImode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, used = 0, 
  frame_related = 0, return_val = 0, u = {fld = {{rt_int = -1008527584, rt_uint = 3286439712, 
        rt_str = 0x3204c3e31720 "\"", rt_rtx = 0x3204c3e31720, rt_rtvec = 0x3204c3e31720, 
        rt_type = 3286439712, rt_addr_diff_vec_flags = {min_align = 32, base_after_vec = 1, 
          min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 1, 
          offset_unsigned = 0, scale = 227}, rt_cselib = 0x3204c3e31720, rt_bit = 0x3204c3e31720, 
        rt_tree = 0x3204c3e31720, rt_bb = 0x3204c3e31720, rt_mem = 0x3204c3e31720, 
        rt_reg = 0x3204c3e31720}}, hwint = {54996047697696}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102701241, sig = {54996046949264, 0, 54996046836352}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3e317a0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$8 = {code = SET, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -1008527648, rt_uint = 3286439648, 
        rt_str = 0x3204c3e316e0 "\"", rt_rtx = 0x3204c3e316e0, rt_rtvec = 0x3204c3e316e0, 
        rt_type = 3286439648, rt_addr_diff_vec_flags = {min_align = 224, base_after_vec = 0, 
          min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 1, 
          offset_unsigned = 0, scale = 227}, rt_cselib = 0x3204c3e316e0, rt_bit = 0x3204c3e316e0, 
        rt_tree = 0x3204c3e316e0, rt_bb = 0x3204c3e316e0, rt_mem = 0x3204c3e316e0, 
        rt_reg = 0x3204c3e316e0}}, hwint = {54996047697632}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102701239, sig = {54996047697760, 0, 0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3d5f2d0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$9 = {code = INSN, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = 9, rt_uint = 9, 
        rt_str = 0x9 <Address 0x9 out of bounds>, rt_rtx = 0x9, rt_rtvec = 0x9, rt_type = TImode, 
        rt_addr_diff_vec_flags = {min_align = 9, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0x9, 
        rt_bit = 0x9, rt_tree = 0x9, rt_bb = 0x9, rt_mem = 0x9, rt_reg = 0x9}}, hwint = {9}, rv = {cl = 1, 
      sign = 0, signalling = 1, canonical = 0, uexp = 0, sig = {0, 0, 0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3d5f280)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$10 = {code = INSN, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = 8, rt_uint = 8, 
        rt_str = 0x8 <Address 0x8 out of bounds>, rt_rtx = 0x8, rt_rtvec = 0x8, rt_type = DImode, 
        rt_addr_diff_vec_flags = {min_align = 8, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0x8, 
        rt_bit = 0x8, rt_tree = 0x8, rt_bb = 0x8, rt_mem = 0x8, rt_reg = 0x8}}, hwint = {8}, rv = {cl = 0, 
      sign = 0, signalling = 1, canonical = 0, uexp = 0, sig = {0, 54996046836432, 0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3e317e0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$11 = {code = SET, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -1008527872, rt_uint = 3286439424, 
        rt_str = 0x3204c3e31600 "'", rt_rtx = 0x3204c3e31600, rt_rtvec = 0x3204c3e31600, 
        rt_type = 3286439424, rt_addr_diff_vec_flags = {min_align = 0, base_after_vec = 0, 
          min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 1, 
          offset_unsigned = 0, scale = 227}, rt_cselib = 0x3204c3e31600, rt_bit = 0x3204c3e31600, 
        rt_tree = 0x3204c3e31600, rt_bb = 0x3204c3e31600, rt_mem = 0x3204c3e31600, 
        rt_reg = 0x3204c3e31600}}, hwint = {54996047697408}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102701232, sig = {54996047697632, 0, 0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3d5f280)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$12 = {code = INSN, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = 8, rt_uint = 8, 
        rt_str = 0x8 <Address 0x8 out of bounds>, rt_rtx = 0x8, rt_rtvec = 0x8, rt_type = DImode, 
        rt_addr_diff_vec_flags = {min_align = 8, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0x8, 
        rt_bit = 0x8, rt_tree = 0x8, rt_bb = 0x8, rt_mem = 0x8, rt_reg = 0x8}}, hwint = {8}, rv = {cl = 0, 
      sign = 0, signalling = 1, canonical = 0, uexp = 0, sig = {0, 54996046836432, 0}}}}
(gdb) 
Continuing.

Breakpoint 1, emit_move_sequence (operands=0x70c27856c780, mode=SImode, scratch_reg=0x0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/config/pa/pa.c:1310
1310	  register rtx operand0 = operands[0];
$13 = {code = REG, mode = SImode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, used = 0, 
  frame_related = 0, return_val = 1, u = {fld = {{rt_int = 28, rt_uint = 28, 
        rt_str = 0x1c <Address 0x1c out of bounds>, rt_rtx = 0x1c, rt_rtvec = 0x1c, rt_type = 28, 
        rt_addr_diff_vec_flags = {min_align = 28, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0x1c, 
        rt_bit = 0x1c, rt_tree = 0x1c, rt_bb = 0x1c, rt_mem = 0x1c, rt_reg = 0x1c}}, hwint = {28}, rv = {
      cl = 0, sign = 1, signalling = 1, canonical = 1, uexp = 0, sig = {28, 0, 54996047741248}}}}
$14 = {code = REG, mode = SImode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, used = 0, 
  frame_related = 0, return_val = 0, u = {fld = {{rt_int = 94, rt_uint = 94, 
        rt_str = 0x5e <Address 0x5e out of bounds>, rt_rtx = 0x5e, rt_rtvec = 0x5e, rt_type = 94, 
        rt_addr_diff_vec_flags = {min_align = 94, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0x5e, 
        rt_bit = 0x5e, rt_tree = 0x5e, rt_bb = 0x5e, rt_mem = 0x5e, rt_reg = 0x5e}}, hwint = {94}, rv = {
      cl = 2, sign = 1, signalling = 1, canonical = 1, uexp = 2, sig = {94, 54996046949216, 458786}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3e31800)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$15 = {code = SET, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -1008527712, rt_uint = 3286439584, 
        rt_str = 0x3204c3e316a0 "\"", rt_rtx = 0x3204c3e316a0, rt_rtvec = 0x3204c3e316a0, 
        rt_type = 3286439584, rt_addr_diff_vec_flags = {min_align = 160, base_after_vec = 0, 
          min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 1, 
          offset_unsigned = 0, scale = 227}, rt_cselib = 0x3204c3e316a0, rt_bit = 0x3204c3e316a0, 
        rt_tree = 0x3204c3e316a0, rt_bb = 0x3204c3e316a0, rt_mem = 0x3204c3e316a0, 
        rt_reg = 0x3204c3e316a0}}, hwint = {54996047697568}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102701237, sig = {54996047697504, 0, 0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3d5f370)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$16 = {code = INSN, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = 14, rt_uint = 14, 
        rt_str = 0xe <Address 0xe out of bounds>, rt_rtx = 0xe, rt_rtvec = 0xe, rt_type = CHImode, 
        rt_addr_diff_vec_flags = {min_align = 14, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0xe, 
        rt_bit = 0xe, rt_tree = 0xe, rt_bb = 0xe, rt_mem = 0xe, rt_reg = 0xe}}, hwint = {14}, rv = {cl = 2, 
      sign = 1, signalling = 1, canonical = 0, uexp = 0, sig = {0, 0, 0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3d7aba0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$17 = {code = CLOBBER, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -1008527712, rt_uint = 3286439584, 
        rt_str = 0x3204c3e316a0 "\"", rt_rtx = 0x3204c3e316a0, rt_rtvec = 0x3204c3e316a0, 
        rt_type = 3286439584, rt_addr_diff_vec_flags = {min_align = 160, base_after_vec = 0, 
          min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 1, 
          offset_unsigned = 0, scale = 227}, rt_cselib = 0x3204c3e316a0, rt_bit = 0x3204c3e316a0, 
        rt_tree = 0x3204c3e316a0, rt_bb = 0x3204c3e316a0, rt_mem = 0x3204c3e316a0, 
        rt_reg = 0x3204c3e316a0}}, hwint = {54996047697568}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102701237, sig = {0, 0, 0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3d7abb0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$18 = {code = CLOBBER, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -1008527776, rt_uint = 3286439520, 
        rt_str = 0x3204c3e31660 "\"", rt_rtx = 0x3204c3e31660, rt_rtvec = 0x3204c3e31660, 
        rt_type = 3286439520, rt_addr_diff_vec_flags = {min_align = 96, base_after_vec = 0, 
          min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 1, 
          offset_unsigned = 0, scale = 227}, rt_cselib = 0x3204c3e31660, rt_bit = 0x3204c3e31660, 
        rt_tree = 0x3204c3e31660, rt_bb = 0x3204c3e31660, rt_mem = 0x3204c3e31660, 
        rt_reg = 0x3204c3e31660}}, hwint = {54996047697504}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102701235, sig = {0, 0, 0}}}}
(gdb) 
Continuing.

Breakpoint 3, emit_insn_after (pattern=0x3204c3d5f3c0, after=0x3204c3e3c400)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4260
4260	  if (INSN_P (after))
$19 = {code = INSN, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = 15, rt_uint = 15, 
        rt_str = 0xf <Address 0xf out of bounds>, rt_rtx = 0xf, rt_rtvec = 0xf, rt_type = CSImode, 
        rt_addr_diff_vec_flags = {min_align = 15, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0xf, 
        rt_bit = 0xf, rt_tree = 0xf, rt_bb = 0xf, rt_mem = 0xf, rt_reg = 0xf}}, hwint = {15}, rv = {cl = 3, 
      sign = 1, signalling = 1, canonical = 0, uexp = 0, sig = {0, 54996046836752, 0}}}}
$20 = {code = NOTE, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = 11, rt_uint = 11, 
        rt_str = 0xb <Address 0xb out of bounds>, rt_rtx = 0xb, rt_rtvec = 0xb, rt_type = DFmode, 
        rt_addr_diff_vec_flags = {min_align = 11, base_after_vec = 0, min_after_vec = 0, max_after_vec = 0, 
          min_after_base = 0, max_after_base = 0, offset_unsigned = 0, scale = 0}, rt_cselib = 0xb, 
        rt_bit = 0xb, rt_tree = 0xb, rt_bb = 0xb, rt_mem = 0xb, rt_reg = 0xb}}, hwint = {11}, rv = {cl = 3, 
      sign = 0, signalling = 1, canonical = 0, uexp = 0, sig = {54996046836512, 54996046836272, 0}}}}
(gdb) 
Continuing.

Breakpoint 2, emit_insn (x=0x3204c3d7abc0)
    at /usr/local/src/gnu/gcc/branches/gcc-4_1-branch/gcc/emit-rtl.c:4433
4433	  rtx last = last_insn;
$21 = {code = USE, mode = VOIDmode, jump = 0, call = 0, unchanging = 0, volatil = 0, in_struct = 0, 
  used = 0, frame_related = 0, return_val = 0, u = {fld = {{rt_int = -1008527712, rt_uint = 3286439584, 
        rt_str = 0x3204c3e316a0 "\"", rt_rtx = 0x3204c3e316a0, rt_rtvec = 0x3204c3e316a0, 
        rt_type = 3286439584, rt_addr_diff_vec_flags = {min_align = 160, base_after_vec = 0, 
          min_after_vec = 1, max_after_vec = 1, min_after_base = 0, max_after_base = 1, 
          offset_unsigned = 0, scale = 227}, rt_cselib = 0x3204c3e316a0, rt_bit = 0x3204c3e316a0, 
        rt_tree = 0x3204c3e316a0, rt_bb = 0x3204c3e316a0, rt_mem = 0x3204c3e316a0, 
        rt_reg = 0x3204c3e316a0}}, hwint = {54996047697568}, rv = {cl = 0, sign = 0, signalling = 0, 
      canonical = 0, uexp = 102701237, sig = {0, 0, 0}}}}
(gdb) 
Continuing.
../seed48_r.i: In function 'main':
../seed48_r.i:1: error: unrecognizable insn:
(insn 9 8 10 1 (set (reg:DI 95)
        (plus:DI (reg:DI 96)
            (const_int -6547 [0xffffffffffffe66d]))) -1 (nil)
    (expr_list:REG_EQUAL (const_int 25214903917 [0x5deece66d])
        (nil)))
../seed48_r.i:1: internal compiler error: in extract_insn, at recog.c:2084
Please submit a full bug report,
with preprocessed source if appropriate.
See <URL:http://gcc.gnu.org/bugs.html> for instructions.

Program exited with code 01.
(gdb) 

[-- Attachment #3: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] gcc ICE PR29114 (was: 64-bit long double support.)

[Mike Frysinger]

[-- Attachment #1.1: Type: text/plain, Size: 170 bytes --]

On Friday 17 November 2006 16:12, Mike Frysinger wrote:
> updated trace with automatic pretty prints attached to each breakpoint

here's the .gdbinit i'm using btw
-mike

[-- Attachment #1.2: Type: application/pgp-signature, Size: 827 bytes --]

[-- Attachment #2: .gdbinit --]
[-- Type: text/plain, Size: 331 bytes --]

file gcc/cc1
set args -fpreprocessed ../seed48_r.i -quiet -dumpbase seed48_r.i -auxbase seed48_r -version -o /tmp/cc9E3kFl.s

b emit_move_sequence
b emit_insn
b emit_insn_after
b emit_insn_before
commands 1
	print *operands[0]
	print *operands[1]
end
commands 2
	if x > 0
		print *x
	end
end
commands 3
	p *pattern
	p *after
end
r

[-- Attachment #3: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux

Re: [parisc-linux] gcc ICE PR29114 (was: 64-bit long double support.)

[Mike Frysinger]

[-- Attachment #1.1: Type: text/plain, Size: 278 bytes --]

On Friday 17 November 2006 16:49, John David Anglin wrote:
> Would you generate a backtrace for this breakpoint?

ive gotta take off for a while, so i've generated a core at that break point 
and uploaded it here:
http://dev.gentoo.org/~vapier/hppa/
hope it helps
-mike

[-- Attachment #1.2: Type: application/pgp-signature, Size: 827 bytes --]

[-- Attachment #2: Type: text/plain, Size: 169 bytes --]

_______________________________________________
parisc-linux mailing list
parisc-linux@lists.parisc-linux.org
http://lists.parisc-linux.org/mailman/listinfo/parisc-linux