[Bug bootstrap/41395] New: [4.5 regression] Bootstrap failure

[Bug bootstrap/41395] New: [4.5 regression] Bootstrap failure

[hjl dot tools at gmail dot com]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1794 bytes --]

On Linux/ia32, revision 151829 gave

[hjl@gnu-9 libgcc]$ /export/build/gnu/gcc/build-i686-linux/./gcc/xgcc
-B/export/build/gnu/gcc/build-i686-linux/./gcc/
-B/usr/gcc-4.5/i686-pc-linux-gnu/bin/ -B/usr/gcc-4.5/i686-pc-linux-gnu/lib/
-isystem /usr/gcc-4.5/i686-pc-linux-gnu/include -isystem
/usr/gcc-4.5/i686-pc-linux-gnu/sys-include    -g -O2 -O2  -g -O2 -DIN_GCC   -W
-Wall -Wwrite-strings -Wcast-qual -Wstrict-prototypes -Wmissing-prototypes
-Wold-style-definition  -isystem ./include  -fPIC -g -DHAVE_GTHR_DEFAULT
-DIN_LIBGCC2 -D__GCC_FLOAT_NOT_NEEDED   -I. -I. -I../.././gcc
-I/net/gnu-13/export/gnu/src/gcc/gcc/libgcc
-I/net/gnu-13/export/gnu/src/gcc/gcc/libgcc/.
-I/net/gnu-13/export/gnu/src/gcc/gcc/libgcc/../gcc
-I/net/gnu-13/export/gnu/src/gcc/gcc/libgcc/../include
-I/net/gnu-13/export/gnu/src/gcc/gcc/libgcc/config/libbid
-DENABLE_DECIMAL_BID_FORMAT -DHAVE_CC_TLS -DUSE_TLS -o _powisf2.o -MT
_powisf2.o -MD -MP -MF _powisf2.dep -DL_powisf2 -c
/net/gnu-13/export/gnu/src/gcc/gcc/libgcc/../gcc/libgcc2.c  
-fvisibility=hidden -DHIDE_EXPORTS
/net/gnu-13/export/gnu/src/gcc/gcc/libgcc/../gcc/libgcc2.c: In function
â__powisf2â:
/net/gnu-13/export/gnu/src/gcc/gcc/libgcc/../gcc/libgcc2.c:1739:1: internal
compiler error: in convert_regs_1, at reg-stack.c:3052
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
[hjl@gnu-9 libgcc]$


-- 
           Summary: [4.5 regression] Bootstrap failure
           Product: gcc
           Version: 4.5.0
            Status: UNCONFIRMED
          Severity: normal
          Priority: P3
         Component: bootstrap
        AssignedTo: unassigned at gcc dot gnu dot org
        ReportedBy: hjl dot tools at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Bootstrap failure

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Bootstrap failure

[hjl dot tools at gmail dot com]

------- Comment #1 from hjl dot tools at gmail dot com  2009-09-18 04:39 -------
Gcc is configured with

 --enable-clocale=gnu --with-system-zlib --enable-checking=assert
--with-demangler-in-ld  --enable-shared --enable-threads=posix --enable-haifa
--prefix=/usr/gcc-4.5 --with-local-prefix=/usr/local

All debug info are corrupted:

GNU gdb (GDB) 7.0.50.20090916-cvs
Copyright (C) 2009 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "i686-pc-linux-gnu".
For bug reporting instructions, please see:
<http://www.gnu.org/software/gdb/bugs/>...
Reading symbols from /export/build/gnu/gcc/build-i686-linux/gcc/cc1...done.
Breakpoint 1 at 0x817bc90
Breakpoint 2 at 0x817c4b0
Breakpoint 3 at 0x80af0a4
Breakpoint 4 at 0x80ae954
(gdb)


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|4.5.0                       |---


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Bootstrap failure

[pinskia at gcc dot gnu dot org]

-- 

pinskia at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
   Target Milestone|---                         |4.5.0


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[hjl dot tools at gmail dot com]

------- Comment #2 from hjl dot tools at gmail dot com  2009-09-18 05:44 -------
It is caused by revision 151800:

http://gcc.gnu.org/ml/gcc-cvs/2009-09/msg00547.html

It is reproducible on both Linux/ia32 and Linux/x86-64.


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |jamborm at gcc dot gnu dot
                   |                            |org
            Summary|[4.5 regression] Bootstrap  |[4.5 regression] Revision
                   |failure                     |151800 failed bootstrap


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[janus at gcc dot gnu dot org]

------- Comment #3 from janus at gcc dot gnu dot org  2009-09-18 10:47 -------
With r151837 the bootstrap works again, but the testsuite still shows a large
number of failures, already with check-gfortran (which was clean recently):

                === gfortran Summary ===

# of expected passes            32101
# of unexpected failures        169
# of expected failures          21
# of unresolved testcases       82
# of unsupported tests          52
/home/jweil/gcc45/build/gcc/testsuite/gfortran/../../gfortran  version 4.5.0
20090918 (experimental) [trunk revision 151837] (GCC) 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #4 from jamborm at gcc dot gnu dot org  2009-09-18 14:42 -------
I can reproduce the problem. --enable-checking=assert is the key configure
option.
Thus, this bug is mine.


-- 

jamborm at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|unassigned at gcc dot gnu   |jamborm at gcc dot gnu dot
                   |dot org                     |org
             Status|UNCONFIRMED                 |ASSIGNED
     Ever Confirmed|0                           |1
   Last reconfirmed|0000-00-00 00:00:00         |2009-09-18 14:42:00
               date|                            |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[janus at gcc dot gnu dot org]

------- Comment #5 from janus at gcc dot gnu dot org  2009-09-18 15:10 -------
(In reply to comment #4)
> --enable-checking=assert is the key configure option.

Are you sure about that? For me, configuring with --enable-checking=no still
yields loads of regressions in the Fortran testsuite. Other configure options I
use are --enable-languages=c,fortran --disable-multilib --disable-bootstrap.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #6 from ubizjak at gmail dot com  2009-09-18 15:29 -------
This revision also fails to bootstrap on alphaev68-unknown-linux-gnu (revision
151799 is OK). stage2 compiler segfaults trying to build libgcc:

configure:3232: /home/uros/gcc-build/./gcc/xgcc -B/home/uros/gcc-build/./gcc/
-B
/usr/local/alphaev68-unknown-linux-gnu/bin/
-B/usr/local/alphaev68-unknown-linux
-gnu/lib/ -isystem /usr/local/alphaev68-unknown-linux-gnu/include -isystem
/usr/
local/alphaev68-unknown-linux-gnu/sys-include    -c -g -O2 -mieee  conftest.c
>&
5
conftest.c:1:1: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.

In fact, stage2 cc1 segfaults on everything:

gdb ~/gcc-build/gcc/cc1
GNU gdb 6.8
Copyright (C) 2008 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.  Type "show copying"
and "show warranty" for details.
This GDB was configured as "alpha-unknown-linux-gnu"...
(no debugging symbols found)
(gdb) set args hello.c
(gdb) run
Starting program: /space/uros/gcc-build/gcc/cc1 hello.c
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)
(no debugging symbols found)

Program received signal SIGSEGV, Segmentation fault.
0x00000001205e2520 in tree_class_check_failed ()
(gdb) bt
#0  0x00000001205e2520 in tree_class_check_failed ()
#1  0x00000001205e33f4 in type_hash_list ()
#2  0x00000001205e88f0 in build_function_type ()
#3  0x00000001205e8d24 in build_function_type_list ()
#4  0x0000000120639e74 in alpha_init_builtins ()
#5  0x00000001200e06ac in c_common_nodes_and_builtins ()
#6  0x0000000120084dc8 in c_init_decl_processing ()
#7  0x00000001200f6178 in c_objc_common_init ()
#8  0x0000000120442594 in toplev_main ()
#9  0x000000012011af78 in main ()

As mentioned in comment #1, debug info is missing/meaningless.

gcc was built with default configure flags:

~/gcc-build/gcc/xgcc -v    
Using built-in specs.
Target: alphaev68-unknown-linux-gnu
Configured with: ../gcc-svn/trunk/configure --enable-languages=c,c++,fortran
Thread model: posix
gcc version 4.5.0 20090917 (experimental) [trunk revision 151835] (GCC) 

(don't bother with revision number, it was patch-reversed to 151800).


-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |ubizjak at gmail dot com


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #7 from ubizjak at gmail dot com  2009-09-18 15:31 -------
On alpha, bootstrap also fails with --enable-checking=assert.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #8 from jamborm at gcc dot gnu dot org  2009-09-18 15:34 -------
(In reply to comment #5)
> (In reply to comment #4)
> > --enable-checking=assert is the key configure option.
> 
> Are you sure about that? For me, configuring with --enable-checking=no still
> yields loads of regressions

I meant as opposed to --enable-checking=yes.  That makes the bootstrap
problem disappear at revision 151800 for me.  I have not looked at the
later revisions yet (which is the next step now I guess).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #9 from ubizjak at gmail dot com  2009-09-18 16:35 -------
(In reply to comment #7)
> On alpha, bootstrap also fails with --enable-checking=assert.

It also fails with --enable-checking=none.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[developer at sandoe-acoustics dot co dot uk]

------- Comment #10 from developer at sandoe-acoustics dot co dot uk  2009-09-18 19:08 -------
on i686-apple-darwin9 bootstrap fails with a variety of different errors
depending on what --enable-checking=xx is set.  
For 
=yes if fails with a lot of dsymutil crashes.  
=runtime it fails per the #1 example, 
=no ... the bootstrap never complete (at least my patience ran out at 4x
normal).


-- 

developer at sandoe-acoustics dot co dot uk changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |developer at sandoe-
                   |                            |acoustics dot co dot uk


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[hjl dot tools at gmail dot com]

------- Comment #11 from hjl dot tools at gmail dot com  2009-09-19 15:58 -------
*** Bug 41409 has been marked as a duplicate of this bug. ***


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |aanisimov at inbox dot ru


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[hjl dot tools at gmail dot com]

------- Comment #12 from hjl dot tools at gmail dot com  2009-09-19 15:59 -------
*** Bug 41407 has been marked as a duplicate of this bug. ***


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[hjl dot tools at gmail dot com]

------- Comment #13 from hjl dot tools at gmail dot com  2009-09-19 17:50 -------
*** Bug 41410 has been marked as a duplicate of this bug. ***


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap with checking disabled

[rguenth at gcc dot gnu dot org]

-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Priority|P3                          |P1
            Summary|[4.5 regression] Revision   |[4.5 regression] Revision
                   |151800 failed bootstrap     |151800 failed bootstrap with
                   |                            |checking disabled


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #14 from ubizjak at gmail dot com  2009-09-19 21:21 -------
Checking is not a problem here, see Comments #9 and #10.


-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[4.5 regression] Revision   |[4.5 regression] Revision
                   |151800 failed bootstrap with|151800 failed bootstrap
                   |checking disabled           |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[developer at sandoe-acoustics dot co dot uk]

[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 477 bytes --]



------- Comment #15 from developer at sandoe-acoustics dot co dot uk  2009-09-19 22:45 -------
just checked; powerpc-apple-darwin9 [at 151880] this also fails.
looking through the error log there do seem to be a large number of garbage
strings in the informational messages; 
e.g. ../../../gcc-4-5-regtest/libgcc/../gcc/unwind-dw2-fde.h: In function
‘last_fde’:
if that's of any relevance.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[rguenth at gcc dot gnu dot org]

------- Comment #18 from rguenth at gcc dot gnu dot org  2009-09-20 09:20 -------
*** Bug 41413 has been marked as a duplicate of this bug. ***


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[rguenth at gcc dot gnu dot org]

------- Comment #16 from rguenth at gcc dot gnu dot org  2009-09-20 09:19 -------
*** Bug 41414 has been marked as a duplicate of this bug. ***


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |aanisimov at inbox dot ru


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[rguenth at gcc dot gnu dot org]

------- Comment #17 from rguenth at gcc dot gnu dot org  2009-09-20 09:20 -------
*** Bug 41415 has been marked as a duplicate of this bug. ***


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[developer at sandoe-acoustics dot co dot uk]

------- Comment #19 from developer at sandoe-acoustics dot co dot uk  2009-09-20 09:42 -------
applying
http://gcc.gnu.org/ml/gcc-patches/2009-09/msg01274.html
causes i686-apple-darwin9 to fail with the (long long) fault for all
--enable-checking=xx I've tried.
Thus, that fault seems separate (PR41405).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jsg at openbsd dot org]

------- Comment #20 from jsg at openbsd dot org  2009-09-20 13:14 -------
bootstrap also fails on OpenBSD/i386 when it used to work a week ago ie:
gcc version 4.5.0 20090913 (experimental) (GCC)

When stage1 is building libgcc:

configure:3003: $? = 0
configure:2992: /usr/users/jsg/src/obj/./gcc/xgcc
-B/usr/users/jsg/src/obj/./gcc/ -B/usr/gcc/i386-unknown-openbsd4.6/bin/
-B/usr/gcc/i386-unknown-openbsd4.6/lib/ -isystem
/usr/gcc/i386-unknown-openbsd4.6/include -isystem
/usr/gcc/i386-unknown-openbsd4.6/sys-include    -v >&5
Reading specs from /usr/users/jsg/src/obj/./gcc/specs
Target: i386-unknown-openbsd4.6
Configured with: ../gcc/configure --prefix=/usr/gcc/ --enable-languages=c
--program-transform-name='s,^,e,' --disable-nls --disable-checking --with-syst
em-zlib --disable-libmudflap --disable-libgomp --disable-tls
--with-as=/usr/bin/as --with-ld=/usr/bin/ld --with-gnu-ld --with-gnu-as
--enable-threads=po
six --enable-cpp --with-gmp=/usr/local --with-mpfr=/usr/local --enable-shared
Thread model: posix
gcc version 4.5.0 20090920 (experimental) (GCC) 
configure:3003: $? = 0
configure:2992: /usr/users/jsg/src/obj/./gcc/xgcc
-B/usr/users/jsg/src/obj/./gcc/ -B/usr/gcc/i386-unknown-openbsd4.6/bin/
-B/usr/gcc/i386-unknown-openbsd4.6/lib/ -isystem
/usr/gcc/i386-unknown-openbsd4.6/include -isystem
/usr/gcc/i386-unknown-openbsd4.6/sys-include    -V >&5
xgcc: '-V' must come at the start of the command line
configure:3003: $? = 1
configure:2992: /usr/users/jsg/src/obj/./gcc/xgcc
-B/usr/users/jsg/src/obj/./gcc/ -B/usr/gcc/i386-unknown-openbsd4.6/bin/
-B/usr/gcc/i386-unknown-openbsd4.6/lib/ -isystem
/usr/gcc/i386-unknown-openbsd4.6/include -isystem
/usr/gcc/i386-unknown-openbsd4.6/sys-include    -qversion >&5
xgcc: unrecognized option '-qversion'
xgcc: no input files
configure:3003: $? = 1
configure:3019: /usr/users/jsg/src/obj/./gcc/xgcc
-B/usr/users/jsg/src/obj/./gcc/ -B/usr/gcc/i386-unknown-openbsd4.6/bin/
-B/usr/gcc/i386-unknown-openbsd4.6/lib/ -isystem
/usr/gcc/i386-unknown-openbsd4.6/include -isystem
/usr/gcc/i386-unknown-openbsd4.6/sys-include    -o conftest -g -O2   conftest.c
 >&5
conftest.c:1:1: internal compiler error: Segmentation fault
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
configure:3022: $? = 1
configure:3210: checking for suffix of object files
configure:3232: /usr/users/jsg/src/obj/./gcc/xgcc
-B/usr/users/jsg/src/obj/./gcc/ -B/usr/gcc/i386-unknown-openbsd4.6/bin/
-B/usr/gcc/i386-unknown-openbsd4.6/lib/ -isystem
/usr/gcc/i386-unknown-openbsd4.6/include -isystem
/usr/gcc/i386-unknown-openbsd4.6/sys-include    -c -g -O2  conftest.c >&5
conftest.c:1:1: internal compiler error: Segmentation fault


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[dominiq at lps dot ens dot fr]

------- Comment #21 from dominiq at lps dot ens dot fr  2009-09-20 15:36 -------
In reply to comment #20
> bootstrap also fails on OpenBSD/i386 when it used to work a week ago ie:
> gcc version 4.5.0 20090913 (experimental) (GCC)

You may try to revert revision 151815 (see pr41405, the patch is in
http://gcc.gnu.org/ml/gcc-patches/2009-09/msg00220.html).


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[hjl at gcc dot gnu dot org]

------- Comment #22 from hjl at gcc dot gnu dot org  2009-09-20 18:00 -------
Subject: Bug 41395

Author: hjl
Date: Sun Sep 20 17:59:44 2009
New Revision: 151905

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=151905
Log:
2009-09-20  H.J. Lu  <hongjiu.lu@intel.com>

        PR middle-end/41395
        * opts.c (decode_options): Don't turn on flag_ipa_sra for opt2.

Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/opts.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[hjl dot tools at gmail dot com]

------- Comment #23 from hjl dot tools at gmail dot com  2009-09-20 19:37 -------
*** Bug 41417 has been marked as a duplicate of this bug. ***


-- 

hjl dot tools at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |m4rkusxxl at web dot de


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #24 from jamborm at gcc dot gnu dot org  2009-09-21 19:49 -------
OK, so I have finally got to the root of the assert failure in
reg-stack.c described in the bug description.  The file is indeed
miscompiled, and the miscompiled  function is VEC_char_base_replace.
A very short one:

VEC_char_base_replace (struct VEC_char_base * vec_, unsigned int ix_, char
obj_)
{
  char old_obj_;
  unsigned int D.41028;

<bb 2>:
  old_obj__4 = vec__1(D)->vec[ix__3(D)];
  vec__1(D)->vec[ix__3(D)] = obj__5(D);
  return old_obj__4;
}

IPA-SRA runs the expression vec__1(D)->vec[ix__3(D)] through
get_ref_base_and_extent, which unfortunately returns bogus pmax_size
of 8.  The aggregate in question is:

struct VEC_char_base
{
  unsigned num;
  unsigned alloc;
  char vec[1];
};

Obviously the punting condition has failed me (again :-).  I have put
some dumps into the function (on x86_64) to see why and these are the
values after the while loop terminates:

--------------------------------------------------
Finished with exp: *vec__1(D)
bit_offset: 64, maxsize: 8
seen_variable_array_ref: 1, seen_union: 0
bit_offset + maxsize: 72
TREE_INT_CST_LOW (TYPE_SIZE (TREE_TYPE (exp))): 96
------------------------------------------------------------

Since 72 is not equal to 96, the maxsize is not set to -1 even though
it should.  It seems that the underlying cause of this is that while
the size of the array type (asize) is 8 bits, it is accounted for 32
in the size of the whole aggregate.  Whether this is a bug or
get_ref_base_and_extent should somehow take alignments into account, I
don't know.  Thus I'd like to ask Richi and others who know exact
meanings of these sizes for advice.

Thanks and sorry for the gross inconvenience, it wouldn't have occurred
to me that I should do _less_ checking.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[hjl dot tools at gmail dot com]

------- Comment #25 from hjl dot tools at gmail dot com  2009-09-21 19:53 -------
(In reply to comment #24)
> OK, so I have finally got to the root of the assert failure in
> reg-stack.c described in the bug description.  The file is indeed
> miscompiled, and the miscompiled  function is VEC_char_base_replace.
> A very short one:
> 

There are 2 problems caused by IPA-SRA:

1. reg-stack.c is miscompiled.
2. DWARF debug info is all messed up.

I don't know if they are related.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[rguenth at gcc dot gnu dot org]

------- Comment #26 from rguenth at gcc dot gnu dot org  2009-09-22 08:21 -------
I will have a look at the get_ref_base_and_extent issue.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[rguenth at gcc dot gnu dot org]

------- Comment #27 from rguenth at gcc dot gnu dot org  2009-09-22 11:48 -------
struct VEC_char_base
{
  unsigned num;
  unsigned alloc;
  short vec[1];
};

short __attribute__((noinline))
foo (struct VEC_char_base *p, int i)
{
  short *q;
  p->vec[i] = 0;
  q = &p->vec[1];
  *q = 1;
  return p->vec[i];
}

extern void abort (void);
extern void *malloc (__SIZE_TYPE__);

int
main()
{
  struct VEC_char_base *p = malloc (sizeof (struct VEC_char_base));
  if (foo (p, 1) != 1)
    abort ();
  return 0;
}


-- 

rguenth at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|jamborm at gcc dot gnu dot  |rguenth at gcc dot gnu dot
                   |org                         |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[rguenth at gcc dot gnu dot org]

------- Comment #28 from rguenth at gcc dot gnu dot org  2009-09-22 14:58 -------
Subject: Bug 41395

Author: rguenth
Date: Tue Sep 22 14:58:05 2009
New Revision: 151981

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=151981
Log:
2009-09-22  Richard Guenther  <rguenther@suse.de>

        PR middle-end/41395
        * tree-dfa.c (get_ref_base_and_extent): Handle trailing
        arrays really properly.

        * gcc.c-torture/execute/pr41395-1.c: New testcase.
        * gcc.c-torture/execute/pr41395-2.c: Likewise.

Added:
    trunk/gcc/testsuite/gcc.c-torture/execute/pr41395-1.c
    trunk/gcc/testsuite/gcc.c-torture/execute/pr41395-2.c
Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/testsuite/ChangeLog
    trunk/gcc/tree-dfa.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #29 from jamborm at gcc dot gnu dot org  2009-09-23 09:28 -------
Thanks. With the patch fixing the problem described in #24, we get
further when compiling with release checking but run into syntax
errors when compiling stage3 libstc++.  And the debug info is still
corrupted.  So I am going to investigate further.


-- 

jamborm at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         AssignedTo|rguenth at gcc dot gnu dot  |jamborm at gcc dot gnu dot
                   |org                         |org


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #30 from ubizjak at gmail dot com  2009-09-23 09:48 -------
(In reply to comment #29)
> Thanks. With the patch fixing the problem described in #24, we get
> further when compiling with release checking but run into syntax
> errors when compiling stage3 libstc++.  And the debug info is still
> corrupted.  So I am going to investigate further.

FYI, I have tried to bootstrap on alpha with patch from comment #22 reverted.
No luck, still fails in the same way.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #31 from jamborm at gcc dot gnu dot org  2009-09-24 22:53 -------
(In reply to comment #29)
> Thanks. With the patch fixing the problem described in #24, we get
> further when compiling with release checking but run into syntax
> errors when compiling stage3 libstc++.

This problem with a reduced testcase is PR 41463.

> And the debug info is still
> corrupted.  So I am going to investigate further.
> 

The debug info may not necessarily be corrupted, it's probably that the old gdb
cannot handle it.  I will try with a new gdb.


-- 

jamborm at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |41463


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #32 from jamborm at gcc dot gnu dot org  2009-09-25 14:43 -------
(In reply to comment #30)
> (In reply to comment #29)
> > Thanks. With the patch fixing the problem described in #24, we get
> > further when compiling with release checking but run into syntax
> > errors when compiling stage3 libstc++.  And the debug info is still
> > corrupted.  So I am going to investigate further.
> 
> FYI, I have tried to bootstrap on alpha with patch from comment #22 reverted.
> No luck, still fails in the same way.
> 

Uros, can you please try it now since we have PR 41463 fixed?  x86_64
with release checking now finally bootstraps so perhaps even alpha
might? :-)

If there are problems, please try to post details as it will be a
different failure from the one seen on intel.

Thanks a lot.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #33 from ubizjak at gmail dot com  2009-09-25 18:09 -------
(In reply to comment #32)

> Uros, can you please try it now since we have PR 41463 fixed?  x86_64
> with release checking now finally bootstraps so perhaps even alpha
> might? :-)
> 
> If there are problems, please try to post details as it will be a
> different failure from the one seen on intel.

No, it still fails... I'll investigate this - any suspects?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #34 from ubizjak at gmail dot com  2009-09-27 12:35 -------
It is tree.o object of stage2 gcc that gets miscompiled when -fipa-sra is added
to BOOT_CFLAGS. If tree.o is substituted with the one from the build without
BOOT_CFLAGS, gcc is again able to compile hello.c without crashing.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #35 from ubizjak at gmail dot com  2009-09-27 14:29 -------
Bingo!

It is build_function_type_list_1 from tree.c that makes problems:

static tree
build_function_type_list_1 (bool vaargs, tree return_type, va_list argp)

This probably makes alpha specific bootstrap failure duplicate of PR41089.  At
least it explains, why other targets don't see the problem.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #36 from ubizjak at gmail dot com  2009-09-27 16:11 -------
This band-aid patch enables bootstrap with patch from comment #22 reverted to
proceed a bit further:

Index: tree.c
===================================================================
--- tree.c      (revision 152218)
+++ tree.c      (working copy)
@@ -7180,7 +7180,9 @@
    function. If VAARGS is set, no void_type_node is appended to the
    the list. ARGP muse be alway be terminated be a NULL_TREE.  */

-static tree
+tree build_function_type_list_1 (bool, tree, va_list)
__attribute__((noinline));
+
+tree
 build_function_type_list_1 (bool vaargs, tree return_type, va_list argp)
 {
   tree t, args, last;

However, it soon trips on a similar problem when building __mulvdi3. I don't
feel like chasing all va_list usages through the source of gcc, and since
patched gcc compiles hello.c successfully, this proves the problem.

So this bug heavily depends on PR41089 ("blocker") as far as alpha is
concerned.


-- 

ubizjak at gmail dot com changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|                            |41089


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #37 from jamborm at gcc dot gnu dot org  2009-09-29 16:16 -------
(In reply to comment #34)
> It is tree.o object of stage2 gcc that gets miscompiled when -fipa-sra is added
> to BOOT_CFLAGS. If tree.o is substituted with the one from the build without
> BOOT_CFLAGS, gcc is again able to compile hello.c without crashing.
> 

Can you please attach pre-processed source of it so that I can try
running it through a cross-compiler?  

Additionally, -fdump-tree-cddce-slim and -fdump-tree-eipa_sra-details
dumps would also be appreciated.  Thanks.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #38 from ubizjak at gmail dot com  2009-09-29 18:36 -------
(In reply to comment #37)

> Can you please attach pre-processed source of it so that I can try
> running it through a cross-compiler?  
> 
> Additionally, -fdump-tree-cddce-slim and -fdump-tree-eipa_sra-details
> dumps would also be appreciated.  Thanks.

I belive that this failure is due to PR41089 and manifest itself in

FAIL: gcc.c-torture/execute/stdarg-1.c execution,  -O3 -fomit-frame-pointer 
FAIL: gcc.c-torture/execute/stdarg-1.c execution,  -O3 -g 

Jakub has promised to look into stdarg issue, so perhaps we should wait a
bit...


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[rth at gcc dot gnu dot org]

------- Comment #39 from rth at gcc dot gnu dot org  2009-09-29 20:54 -------
(In reply to comment #35)
> static tree
> build_function_type_list_1 (bool vaargs, tree return_type, va_list argp)

Passing va_list by value is non-portable.  One fix here is to pass argp
by reference.  Dunno if that fixes this problem though, since the main
problem for Alpha seems to be the stdarg thing.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #40 from ubizjak at gmail dot com  2009-09-30 08:12 -------
(In reply to comment #39)

> > static tree
> > build_function_type_list_1 (bool vaargs, tree return_type, va_list argp)
> 
> Passing va_list by value is non-portable.  One fix here is to pass argp
> by reference.  Dunno if that fixes this problem though, since the main
> problem for Alpha seems to be the stdarg thing.

Regarding non-portability of va-arg, I have found an interesting explanation at
[1] which I qoute here for reference:

<quote>
If you go on to more complex cases than asked and shown, one thing
that is *not* well defined is using the same va_list in the caller
after the callee returns. In particular, it is not specified by the
Standard whether va_list is an array type and hence passed "by
reference" (that is, by decayed pointer) and shared, or a type like a
pointer or struct that is passed by value and not shared. If this is
an issue, pass an explicit pointer (and dereference) to force sharing,
or in C99 only use va_copy and a second va_list, in either the caller
or callee, to prevent it.
</quote>

I don't think that sharing is the problem in this particular case, since we
just call va_end after va_list is passed to the function.

[1] http://www.velocityreviews.com/forums/t317863-valist-usage.html


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #41 from jamborm at gcc dot gnu dot org  2009-09-30 11:35 -------
(In reply to comment #38)
> (In reply to comment #37)
> 
> > Can you please attach pre-processed source of it so that I can try
> > running it through a cross-compiler?  
> > 
> > Additionally, -fdump-tree-cddce-slim and -fdump-tree-eipa_sra-details
> > dumps would also be appreciated.  Thanks.
> 
> I belive that this failure is due to PR41089 and manifest itself in
> 

Perhaps, the dumps would allow me to make sure.  But if it's various
va_lists failing, then it is quite probable.

> Jakub has promised to look into stdarg issue, so perhaps we should wait a
> bit...
> 

I do not want to wait with enabling IPA-SRA by default for too long
(the fact that I probably won't make it before the LTO merge is bad
enough).  Thus I have resorted to disabling their IPA-scalarization.
I will attach a patch, can you please try it?


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #42 from jamborm at gcc dot gnu dot org  2009-09-30 11:37 -------
Created an attachment (id=18676)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18676&action=view)
Making IPA-SRA ignore va_lists

This patch prevents IPA-SRA from considering va_list structs as candidates for
processing.  (Already bootstrapped and tested on x86_64 but needs testing on
alpha). 


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[ubizjak at gmail dot com]

------- Comment #43 from ubizjak at gmail dot com  2009-09-30 13:34 -------
(In reply to comment #42)
> Created an attachment (id=18676)
 --> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=18676&action=view) [edit]
> Making IPA-SRA ignore va_lists
> 
> This patch prevents IPA-SRA from considering va_list structs as candidates for
> processing.  (Already bootstrapped and tested on x86_64 but needs testing on
> alpha). 

Yes, this patch bootstrapped OK (with recent opts.c change reverted).
Currently, regression tests are running, they will finish in a couple of hours.

The patch is OK as far as alpha is concerned.


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #44 from jamborm at gcc dot gnu dot org  2009-10-01 11:30 -------
Subject: Bug 41395

Author: jamborm
Date: Thu Oct  1 11:30:12 2009
New Revision: 152366

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=152366
Log:
2009-10-01  Martin Jambor  <mjambor@suse.cz>

        PR bootstrap/41395
        * tree-sra.c (is_va_list_type): New function.
        (find_var_candidates): Call is_va_list_type.
        (find_param_candidates): Check that the type or the type pointed
        to are not va_list types.



Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/tree-sra.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #45 from jamborm at gcc dot gnu dot org  2009-10-01 14:47 -------
Right, so I belieive all problems that were reported here (and were
indeed relevant to IPA-SRA) are now dealt with.  x86_64 and i386
bootstraps and checks nicely with both "yes" and "release" checking,
alpha bootstraps too and I also tested IA64.  As far as I can tell,
bootstrapped gcc has debug info in order.  Thus I will ask Richi when
to enable IPA-SRA by default at -O2.

Thanks everyone for your cooperation and patience.


-- 

jamborm at gcc dot gnu dot org changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  BugsThisDependsOn|41089, 41463                |


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #46 from jamborm at gcc dot gnu dot org  2009-10-06 13:31 -------
Subject: Bug 41395

Author: jamborm
Date: Tue Oct  6 13:31:40 2009
New Revision: 152492

URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=152492
Log:
2009-10-06  Martin Jambor  <mjambor@suse.cz>

        PR bootstrap/41395
        * opts.c (decode_options): Run IPA-SRA at -O2.


Modified:
    trunk/gcc/ChangeLog
    trunk/gcc/opts.c


-- 


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395

[Bug bootstrap/41395] [4.5 regression] Revision 151800 failed bootstrap

[jamborm at gcc dot gnu dot org]

------- Comment #47 from jamborm at gcc dot gnu dot org  2009-10-06 14:11 -------
Finally, all problems that have cause this havoc are fixed.


-- 

jamborm at gcc dot gnu dot org changed:

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


http://gcc.gnu.org/bugzilla/show_bug.cgi?id=41395