* 4.6.1 bootstrap failure on ia64
@ 2011-10-14 9:00 Salvatore Filippone
2011-10-14 11:39 ` Alexander Monakov
2011-10-15 3:21 ` Ian Lance Taylor
0 siblings, 2 replies; 8+ messages in thread
From: Salvatore Filippone @ 2011-10-14 9:00 UTC (permalink / raw)
To: gcc-help
Hello,
I have access to a relatively old Itanium cluster, which I do not
control, and am trying to build 4.6 (0 and 1) as well as the current
trunk, but I keep getting bootstrap failures; I am stuck with a snapshot
of trunk from last January, some time after that it just stopped
building.
This is the error message:
===============================================================
make[2]: Entering directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
make[3]: Entering directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
rm -f stage_current
make[3]: Leaving directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
Comparing stages 2 and 3
warning: gcc/cc1-checksum.o differs
warning: gcc/cc1plus-checksum.o differs
Bootstrap comparison failure!
gcc/insn-emit.o differs
gcc/sel-sched-ir.o differs
make[2]: *** [compare] Error 1
make[2]: Leaving directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
make[1]: *** [stage3-bubble] Error 2
make[1]: Leaving directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
make: *** [bootstrap] Error 2
[pasqua@n64 obj-4.6.1]$ uname -a
Linux n64 2.4.21-15.11hp.XCsmp #1 SMP Wed Dec 22 09:45:49 EST 2004 ia64 ia64 ia64 GNU/Linux
================================================================
Any suggestions as to what to try next?
Thanks a lot
Salvatore
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: 4.6.1 bootstrap failure on ia64
2011-10-14 9:00 4.6.1 bootstrap failure on ia64 Salvatore Filippone
@ 2011-10-14 11:39 ` Alexander Monakov
2011-10-15 3:21 ` Ian Lance Taylor
1 sibling, 0 replies; 8+ messages in thread
From: Alexander Monakov @ 2011-10-14 11:39 UTC (permalink / raw)
To: Salvatore Filippone; +Cc: gcc-help
On Fri, 14 Oct 2011, Salvatore Filippone wrote:
> Hello,
> I have access to a relatively old Itanium cluster, which I do not
> control, and am trying to build 4.6 (0 and 1) as well as the current
> trunk, but I keep getting bootstrap failures; I am stuck with a snapshot
> of trunk from last January, some time after that it just stopped
> building.
It's possible that your system compiler is too old and it miscompiles stage1.
I would suggest to first build a slightly older GCC (e.g. from 4.5 branch) and
then use that compiler to build trunk or 4.6.
Alexander
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: 4.6.1 bootstrap failure on ia64
2011-10-14 9:00 4.6.1 bootstrap failure on ia64 Salvatore Filippone
2011-10-14 11:39 ` Alexander Monakov
@ 2011-10-15 3:21 ` Ian Lance Taylor
2011-10-15 16:01 ` Salvatore Filippone
1 sibling, 1 reply; 8+ messages in thread
From: Ian Lance Taylor @ 2011-10-15 3:21 UTC (permalink / raw)
To: salvatore.filippone; +Cc: gcc-help
Salvatore Filippone <salvatore.filippone@uniroma2.it> writes:
> I have access to a relatively old Itanium cluster, which I do not
> control, and am trying to build 4.6 (0 and 1) as well as the current
> trunk, but I keep getting bootstrap failures; I am stuck with a snapshot
> of trunk from last January, some time after that it just stopped
> building.
> This is the error message:
> ===============================================================
> make[2]: Entering directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
> make[3]: Entering directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
> rm -f stage_current
> make[3]: Leaving directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
> Comparing stages 2 and 3
> warning: gcc/cc1-checksum.o differs
> warning: gcc/cc1plus-checksum.o differs
> Bootstrap comparison failure!
> gcc/insn-emit.o differs
> gcc/sel-sched-ir.o differs
> make[2]: *** [compare] Error 1
> make[2]: Leaving directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
> make[1]: *** [stage3-bubble] Error 2
> make[1]: Leaving directory `/home3/pasqua/GNUBUILD/obj-4.6.1'
> make: *** [bootstrap] Error 2
> [pasqua@n64 obj-4.6.1]$ uname -a
> Linux n64 2.4.21-15.11hp.XCsmp #1 SMP Wed Dec 22 09:45:49 EST 2004 ia64 ia64 ia64 GNU/Linux
> ================================================================
>
> Any suggestions as to what to try next?
Compare insn-emit.o and sel-sched-ir.o in the stage2/gcc and stage3/gcc
directories, and find out how they differ.
Ian
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: 4.6.1 bootstrap failure on ia64
2011-10-15 3:21 ` Ian Lance Taylor
@ 2011-10-15 16:01 ` Salvatore Filippone
2011-10-16 2:23 ` Ian Lance Taylor
0 siblings, 1 reply; 8+ messages in thread
From: Salvatore Filippone @ 2011-10-15 16:01 UTC (permalink / raw)
To: Ian Lance Taylor; +Cc: gcc-help
Il giorno ven, 14/10/2011 alle 20.20 -0700, Ian Lance Taylor ha scritto:
> Salvatore Filippone <salvatore.filippone@uniroma2.it> writes:
> >
> > Any suggestions as to what to try next?
>
> Compare insn-emit.o and sel-sched-ir.o in the stage2/gcc and stage3/gcc
> directories, and find out how they differ.
>
> Ian
>
They differ in size, with the stage3 one being significantly larger;
being binary files I wouldn't know how to compare the contents.
Thanks
Salvatore
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: 4.6.1 bootstrap failure on ia64
2011-10-15 16:01 ` Salvatore Filippone
@ 2011-10-16 2:23 ` Ian Lance Taylor
2011-10-16 14:24 ` Salvatore Filippone
2011-10-17 8:34 ` Salvatore Filippone
0 siblings, 2 replies; 8+ messages in thread
From: Ian Lance Taylor @ 2011-10-16 2:23 UTC (permalink / raw)
To: salvatore.filippone; +Cc: gcc-help
Salvatore Filippone <salvatore.filippone@uniroma2.it> writes:
> Il giorno ven, 14/10/2011 alle 20.20 -0700, Ian Lance Taylor ha scritto:
>> Salvatore Filippone <salvatore.filippone@uniroma2.it> writes:
>> >
>> > Any suggestions as to what to try next?
>>
>> Compare insn-emit.o and sel-sched-ir.o in the stage2/gcc and stage3/gcc
>> directories, and find out how they differ.
>>
>> Ian
>>
>
> They differ in size, with the stage3 one being significantly larger;
> being binary files I wouldn't know how to compare the contents.
A good start is running objdump -x on each and comparing the output. If
they seem identical there, try objdump -d.
It's normal for the stage3 .o file to include debugging info and the
stage2 .o file to not include it. The comparison script will ignore
that. So it's normal for the stage3 .o file to be significantly larger.
Ian
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: 4.6.1 bootstrap failure on ia64
2011-10-16 2:23 ` Ian Lance Taylor
@ 2011-10-16 14:24 ` Salvatore Filippone
2011-10-17 8:34 ` Salvatore Filippone
1 sibling, 0 replies; 8+ messages in thread
From: Salvatore Filippone @ 2011-10-16 14:24 UTC (permalink / raw)
To: Ian Lance Taylor; +Cc: gcc-help
Il giorno sab, 15/10/2011 alle 19.23 -0700, Ian Lance Taylor ha scritto:
> It's normal for the stage3 .o file to include debugging info and the
> stage2 .o file to not include it. The comparison script will ignore
> that. So it's normal for the stage3 .o file to be significantly
> larger.
>
> Ian
From a quick look at the output of objdump -x, as far as I understand,
the difference is only debug info.
I just built 4.4.6, and with it I am bootstrapping 4.5.0; we'll see what
happens.
Do I have to install my own binutils? The system binutils are at 2.14.
Thanks
Salvatore
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: 4.6.1 bootstrap failure on ia64
2011-10-16 2:23 ` Ian Lance Taylor
2011-10-16 14:24 ` Salvatore Filippone
@ 2011-10-17 8:34 ` Salvatore Filippone
2011-10-20 9:02 ` Salvatore Filippone
1 sibling, 1 reply; 8+ messages in thread
From: Salvatore Filippone @ 2011-10-17 8:34 UTC (permalink / raw)
To: Ian Lance Taylor; +Cc: gcc-help
Update:
The same kind of error, mismatch in insn-emit.o between stage2 and stage
3, happens in 4.5.0 and in 4.6.1.
Looking at objdump -x it seems to me that the major difference is in the
debug symbols present in stage3. There are a few differing addresses; I
do not know the internals of obj files well enough to be certain, it
looks like they are just related to the inserted debug symbols taking up
space. The start of the comparison is attached below, there is a huge
section in the stage3 file I am not including, but I can provide the
complete dumps if needed. This is from the 4.5.0 boot, the 4.6.1 is very
similar.
Thoughts/workarounds?
Thanks a lot
Salvatore
==============================================================
< stage2-gcc/insn-emit.o: file format elf64-ia64-little
< stage2-gcc/insn-emit.o
---
> stage3-gcc/insn-emit.o: file format elf64-ia64-little
> stage3-gcc/insn-emit.o
17c17,23
< 3 .IA_64.unwind_info 000049e8 0000000000000000 0000000000000000
000346b0 2**3
---
> 3 .debug_abbrev 00000427 0000000000000000 0000000000000000
000346b0 2**0
> CONTENTS, READONLY, DEBUGGING
> 4 .debug_info 0001920d 0000000000000000 0000000000000000
00034ad7 2**0
> CONTENTS, RELOC, READONLY, DEBUGGING
> 5 .debug_line 0000aa61 0000000000000000 0000000000000000
0004dce4 2**0
> CONTENTS, RELOC, READONLY, DEBUGGING
> 6 .IA_64.unwind_info 000049e8 0000000000000000 0000000000000000
00058748 2**3
19c25
< 4 .IA_64.unwind 00003b70 0000000000000000 0000000000000000
00039098 2**3
---
> 7 .IA_64.unwind 00003b70 0000000000000000 0000000000000000
0005d130 2**3
21c27
< 5 .rodata.str1.8 000000d0 0000000000000000 0000000000000000
0003cc08 2**3
---
> 8 .rodata.str1.8 000000d0 0000000000000000 0000000000000000
00060ca0 2**3
23c29
< 6 .sdata 00000020 0000000000000000 0000000000000000
0003ccd8 2**3
---
> 9 .sdata 00000020 0000000000000000 0000000000000000
00060d70 2**3
25c31
< 7 .rodata 000000ae 0000000000000000 0000000000000000
0003ccf8 2**3
---
> 10 .rodata 000000ae 0000000000000000 0000000000000000
00060d90 2**3
27c33,45
< 8 .comment 00000012 0000000000000000 0000000000000000
0003cda6 2**0
---
> 11 .debug_loc 000159c7 0000000000000000 0000000000000000
00060e3e 2**0
> CONTENTS, READONLY, DEBUGGING
> 12 .debug_pubnames 00003457 0000000000000000 0000000000000000
00076805 2**0
> CONTENTS, RELOC, READONLY, DEBUGGING
> 13 .debug_pubtypes 00000e7f 0000000000000000 0000000000000000
00079c5c 2**0
> CONTENTS, RELOC, READONLY, DEBUGGING
> 14 .debug_aranges 00000030 0000000000000000 0000000000000000
0007aadb 2**0
> CONTENTS, RELOC, READONLY, DEBUGGING
> 15 .debug_ranges 00004b50 0000000000000000 0000000000000000
0007ab0b 2**0
> CONTENTS, READONLY, DEBUGGING
> 16 .debug_str 0000f5cc 0000000000000000 0000000000000000
0007f65b 2**0
> CONTENTS, READONLY, DEBUGGING
> 17 .comment 00000012 0000000000000000 0000000000000000
0008ec27 2**0
33a52,54
> 0000000000000000 l d .debug_abbrev 0000000000000000
> 0000000000000000 l d .debug_info 0000000000000000
> 0000000000000000 l d .debug_line 0000000000000000
46a68,73
> 0000000000000000 l d .debug_loc 0000000000000000
> 0000000000000000 l d .debug_pubnames 0000000000000000
> 0000000000000000 l d .debug_pubtypes 0000000000000000
> 0000000000000000 l d .debug_aranges 0000000000000000
> 0000000000000000 l d .debug_ranges 0000000000000000
> 0000000000000000 l d .debug_str 0000000000000000
^ permalink raw reply [flat|nested] 8+ messages in thread
* Re: 4.6.1 bootstrap failure on ia64
2011-10-17 8:34 ` Salvatore Filippone
@ 2011-10-20 9:02 ` Salvatore Filippone
0 siblings, 0 replies; 8+ messages in thread
From: Salvatore Filippone @ 2011-10-20 9:02 UTC (permalink / raw)
To: Ian Lance Taylor; +Cc: gcc-help
I have finally been able to bootstrap 4.6.1; to do it I had to build my
own copy of binutils 2.20, perhaps this will be useful to someone else
with the same problem
Regards
Salvatore
^ permalink raw reply [flat|nested] 8+ messages in thread
end of thread, other threads:[~2011-10-20 9:02 UTC | newest]
Thread overview: 8+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2011-10-14 9:00 4.6.1 bootstrap failure on ia64 Salvatore Filippone
2011-10-14 11:39 ` Alexander Monakov
2011-10-15 3:21 ` Ian Lance Taylor
2011-10-15 16:01 ` Salvatore Filippone
2011-10-16 2:23 ` Ian Lance Taylor
2011-10-16 14:24 ` Salvatore Filippone
2011-10-17 8:34 ` Salvatore Filippone
2011-10-20 9:02 ` Salvatore Filippone
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).