--- Comment #34 from jakub at gcc dot gnu dot org 2007-12-15 23:09 ---
Subject: Bug 34003
Author: jakub
Date: Sat Dec 15 23:08:47 2007
New Revision: 130979
URL: http://gcc.gnu.org/viewcvs?root=gccview=revrev=130979
Log:
PR bootstrap/34003
* c-decl.c (merge_decls): Copy
--- Comment #18 from r dot emrich at de dot tecosim dot com 2007-12-13
11:58 ---
Here the complete linker step. I added the -v -t +vallcompatwarnings command
line switches.
/usr/ccs/bin/ld -v -L/lib/pa1.1 -L/usr/lib/pa1.1 -z -u main -u __gcc_plt_call
-o build/genconstants
--- Comment #19 from r dot emrich at de dot tecosim dot com 2007-12-13
12:06 ---
/usr/ccs/bin/nm build/rtl.o
Symbols from build/rtl.o:
NameValue Scope TypeSubspace
$$dyncall | |undef |milli |
$CODE$ | 0|static|data
--- Comment #20 from jakub at gcc dot gnu dot org 2007-12-13 13:06 ---
Can you attach rtl.s and ggc-none.s (just add -save-temps when compiling
rtl.o and ggc-none.o)?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34003
--- Comment #22 from r dot emrich at de dot tecosim dot com 2007-12-13
14:09 ---
Created an attachment (id=14743)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14743action=view)
assembler code
rtl.s
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34003
--- Comment #21 from r dot emrich at de dot tecosim dot com 2007-12-13
14:08 ---
Created an attachment (id=14742)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14742action=view)
assembler code
ggc-none.s
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34003
--- Comment #23 from jakub at gcc dot gnu dot org 2007-12-13 16:53 ---
rtl.s has
.CALL ARGW0=GR
bl ggc_free,%r2
nop
in VEC_rtx_gc_free, but no corresponding .IMPORT ggc_free, CODE
directive. Could you please also attach rtl.i and exact set of options which
were
--- Comment #26 from dave at hiauly1 dot hia dot nrc dot ca 2007-12-13
17:23 ---
Subject: Re: [4.2/4.3 Regression] gcc 4.3.0 unable to bootstrap itself;
Unsatisfied symbols: ggc_free
rtl.s has
.CALL ARGW0=GR
bl ggc_free,%r2
nop
in VEC_rtx_gc_free, but no
--- Comment #27 from jakub at gcc dot gnu dot org 2007-12-13 17:34 ---
First of all I wonder what's the point of the PA ASM_OUTPUT_EXTERNAL_REAL vs.
ASM_OUTPUT_EXTERNAL hacks. The middle-end already does the deferring itself,
see
--- Comment #24 from r dot emrich at de dot tecosim dot com 2007-12-13
17:01 ---
Created an attachment (id=14744)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14744action=view)
preprocessed source
gcc -c -g -fkeep-inline-functions -DIN_GCC -W -Wall -Wwrite-strings
--- Comment #25 from jakub at gcc dot gnu dot org 2007-12-13 17:18 ---
Ok, got it, minimal testcase is:
extern void foo (void);
int bar (void) { foo (); return 1; }
extern void foo (void);
with -fno-unit-at-a-time there is no .IMPORT directive for foo (removing the
prototype after use
--- Comment #28 from jakub at gcc dot gnu dot org 2007-12-13 17:53 ---
Created an attachment (id=14745)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14745action=view)
gcc43-pr34003.patch
Indeed, removing that second deferral cures this.
--
--- Comment #29 from dave at hiauly1 dot hia dot nrc dot ca 2007-12-13
18:54 ---
Subject: Re: [4.2/4.3 Regression] gcc 4.3.0 unable to bootstrap itself;
Unsatisfied symbols: ggc_free
First of all I wonder what's the point of the PA ASM_OUTPUT_EXTERNAL_REAL vs.
ASM_OUTPUT_EXTERNAL
--- Comment #30 from jakub at gcc dot gnu dot org 2007-12-13 18:58 ---
If PA wants to keep the guards (!TREE_ASM_WRITTEN SYMBOL_REF_REFERENCED_P
()),
it can, but that surely can be done in ASM_OUTPUT_EXTERNAL too, doesn't need
any deferral. In that case this will be broken again.
--- Comment #31 from dave at hiauly1 dot hia dot nrc dot ca 2007-12-13
19:19 ---
Subject: Re: [4.2/4.3 Regression] gcc 4.3.0 unable to bootstrap itself;
Unsatisfied symbols: ggc_free
If PA wants to keep the guards (!TREE_ASM_WRITTEN SYMBOL_REF_REFERENCED_P
()),
it can, but that
--- Comment #32 from jakub at gcc dot gnu dot org 2007-12-13 19:37 ---
I have a patch to fix c-decl.c and pa_encode_section_info which together fixes
this bug. I'll prepare a testcase and submit. The removal of the unnecessary
deferral can be postponed for later.
--
--- Comment #33 from jakub at gcc dot gnu dot org 2007-12-13 20:21 ---
Created an attachment (id=14747)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14747action=view)
gcc43-pr34003.patch
I'm testing it on x86_64-linux now to see if it doesn't cause any issues.
--
jakub at gcc
--- Comment #16 from jakub at gcc dot gnu dot org 2007-12-11 11:40 ---
But according to PR32636, gcc can bootstrap on hppa2.0w-hp-hpux11.11.
Perhaps you have a buggy linker? nm dump you posted shows definition of those
symbols...
--
jakub at gcc dot gnu dot org changed:
--- Comment #17 from r dot emrich at de dot tecosim dot com 2007-12-11
12:26 ---
I'm completly lost. On the same machine:
Actual 4.2 branch works fine. Bootstrapping without java and using the
resulting gcc as bootstrap compiler for both configurations
hppa2.0w-hp-hpux11.00 and
--
mmitchel at gcc dot gnu dot org changed:
What|Removed |Added
Priority|P3 |P1
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34003
--- Comment #13 from r dot emrich at de dot tecosim dot com 2007-11-09
10:33 ---
Your right, seems to be platform-specific. hppa64-hp-hpux11.00 is ok.
I will investigate.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34003
--- Comment #14 from r dot emrich at de dot tecosim dot com 2007-11-09
14:40 ---
nm ggc-none.o
Symbols from ggc-none.o:
NameValue Scope TypeSubspace
$CODE$ | 0|static|data |$CODE$
$CODE$ | 0|static|data |$CODE$
--- Comment #15 from r dot emrich at de dot tecosim dot com 2007-11-09
15:06 ---
Created an attachment (id=14515)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14515action=view)
preprocessed source
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34003
--- Comment #12 from rsandifo at gcc dot gnu dot org 2007-11-08 16:10
---
Interesting. Something platform-specific must be going wrong here.
The original build failure shouldn't have happened even without
the attached patch because ggc-none.o should provide a definition
of ggc_free.
--- Comment #11 from r dot emrich at de dot tecosim dot com 2007-11-07
10:58 ---
Created an attachment (id=14494)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14494action=view)
preprocessed source
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=34003
--- Comment #10 from r dot emrich at de dot tecosim dot com 2007-11-07
10:57 ---
Patch solves the problem for rtl.c in gcc-4.3.0, but later:
gcc -g -fkeep-inline-functions -DIN_GCC -W -Wall -Wwrite-strings
-Wstrict-prototypes -Wmissing-prototypes -Wold-style-definition
--- Comment #8 from pinskia at gcc dot gnu dot org 2007-11-06 18:52 ---
Confirmed.
Caused by:
2006-02-18 Richard Sandiford [EMAIL PROTECTED]
--
pinskia at gcc dot gnu dot org changed:
What|Removed |Added
--- Comment #9 from rsandifo at gcc dot gnu dot org 2007-11-06 22:15
---
Created an attachment (id=14490)
-- (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14490action=view)
possible patch
I can't produce the bug on x86_64-linux-gnu, but could you give the
attached patch a go?
--
rsandifo at gcc dot gnu dot org changed:
What|Removed |Added
AssignedTo|unassigned at gcc dot gnu |rsandifo at gcc dot gnu dot
|dot org
--
rsandifo at gcc dot gnu dot org changed:
What|Removed |Added
CC|rsandifo at sources dot |
|redhat dot com |
30 matches
Mail list logo
