* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
@ 2007-11-04 1:38 ` lindevel at gmx dot net
2007-11-04 2:02 ` lindevel at gmx dot net
` (39 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: lindevel at gmx dot net @ 2007-11-04 1:38 UTC (permalink / raw)
To: gcc-bugs
------- Comment #1 from lindevel at gmx dot net 2007-11-04 01:38 -------
(Make summary less cryptic)
--
lindevel at gmx dot net changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|Building libstdc++-v3: |Building libstdc++-v3:
|include/limits: stray |include/limits: stray '\275'
|‘\275’ in |in program
|program |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
2007-11-04 1:38 ` [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' " lindevel at gmx dot net
@ 2007-11-04 2:02 ` lindevel at gmx dot net
2007-11-04 6:31 ` pinskia at gcc dot gnu dot org
` (38 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: lindevel at gmx dot net @ 2007-11-04 2:02 UTC (permalink / raw)
To: gcc-bugs
------- Comment #2 from lindevel at gmx dot net 2007-11-04 02:02 -------
Created an attachment (id=14481)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14481&action=view)
build.log
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
2007-11-04 1:38 ` [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' " lindevel at gmx dot net
2007-11-04 2:02 ` lindevel at gmx dot net
@ 2007-11-04 6:31 ` pinskia at gcc dot gnu dot org
2007-11-04 11:47 ` lindevel at gmx dot net
` (37 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2007-11-04 6:31 UTC (permalink / raw)
To: gcc-bugs
------- Comment #3 from pinskia at gcc dot gnu dot org 2007-11-04 06:31 -------
first don't use -fvisibility-inlines-hidden, it changes the ABI.
Second what happens if unset CFLAGS?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (2 preceding siblings ...)
2007-11-04 6:31 ` pinskia at gcc dot gnu dot org
@ 2007-11-04 11:47 ` lindevel at gmx dot net
2007-11-09 9:50 ` lindevel at gmx dot net
` (36 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: lindevel at gmx dot net @ 2007-11-04 11:47 UTC (permalink / raw)
To: gcc-bugs
------- Comment #4 from lindevel at gmx dot net 2007-11-04 11:47 -------
Created an attachment (id=14482)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=14482&action=view)
build.log ICE
Building without setting CFLAGS or CXXFLAGS results in an ICE:
/var/tmp/portage/sys-devel/gcc-4.3.0_pre20071103/work/gcc-4.3.0-20071103/libiberty/hashtab.c:
In function ‘htab_expand’:
/var/tmp/portage/sys-devel/gcc-4.3.0_pre20071103/work/gcc-4.3.0-20071103/libiberty/hashtab.c:554:
internal compiler error: Segmentation fault
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (3 preceding siblings ...)
2007-11-04 11:47 ` lindevel at gmx dot net
@ 2007-11-09 9:50 ` lindevel at gmx dot net
2007-11-10 6:42 ` dirtyepic at gentoo dot org
` (35 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: lindevel at gmx dot net @ 2007-11-09 9:50 UTC (permalink / raw)
To: gcc-bugs
------- Comment #5 from lindevel at gmx dot net 2007-11-09 09:50 -------
The problem (stray characters in include/limits) persists after I rebuilt the
whole system to wipe out any leftovers from previous attempts at vectorisation
and hidden inlines. Setting LANG and LC_ALL also didn't help. It seems as if
the file itself is wrong...
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (4 preceding siblings ...)
2007-11-09 9:50 ` lindevel at gmx dot net
@ 2007-11-10 6:42 ` dirtyepic at gentoo dot org
2007-11-10 15:57 ` dirtyepic at gentoo dot org
` (34 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: dirtyepic at gentoo dot org @ 2007-11-10 6:42 UTC (permalink / raw)
To: gcc-bugs
------- Comment #6 from dirtyepic at gentoo dot org 2007-11-10 06:42 -------
this error doesn't happen on x86 but i did reproduce it w/ make
profiledbootstrap using today's trunk on x86_64. if no one else is hitting
this it points to a problem on our end, but i'll look further into it soon.
--
dirtyepic at gentoo dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |dirtyepic at gentoo dot org
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (5 preceding siblings ...)
2007-11-10 6:42 ` dirtyepic at gentoo dot org
@ 2007-11-10 15:57 ` dirtyepic at gentoo dot org
2007-11-11 2:42 ` dirtyepic at gentoo dot org
` (33 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: dirtyepic at gentoo dot org @ 2007-11-10 15:57 UTC (permalink / raw)
To: gcc-bugs
------- Comment #7 from dirtyepic at gentoo dot org 2007-11-10 15:56 -------
looks like this only occurs with profiledbootstrap. i'll test a vanilla build
to see if this is Gentoo specific.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (6 preceding siblings ...)
2007-11-10 15:57 ` dirtyepic at gentoo dot org
@ 2007-11-11 2:42 ` dirtyepic at gentoo dot org
2007-12-04 17:43 ` h dot mth at web dot de
` (32 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: dirtyepic at gentoo dot org @ 2007-11-11 2:42 UTC (permalink / raw)
To: gcc-bugs
------- Comment #8 from dirtyepic at gentoo dot org 2007-11-11 02:42 -------
reproduced with gcc-4.3-20071109 snapshot.
dirtyepic@kali ~/tmp/gccbuild $ ../gcc-4.3-20071109/configure --disable-nls
--with-system-zlib --disable-checking --disable-werror --enable-secureplt
--disable-libunwind-exceptions --enable-multilib --enable-languages=c,c++
dirtyepic@kali ~/tmp/gccbuild $ make profiledbootstrap -j1
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (7 preceding siblings ...)
2007-11-11 2:42 ` dirtyepic at gentoo dot org
@ 2007-12-04 17:43 ` h dot mth at web dot de
2007-12-05 13:01 ` h dot mth at web dot de
` (31 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: h dot mth at web dot de @ 2007-12-04 17:43 UTC (permalink / raw)
To: gcc-bugs
[-- Warning: decoded text below may be mangled, UTF-8 assumed --]
[-- Attachment #1: Type: text/plain, Size: 1107 bytes --]
------- Comment #9 from h dot mth at web dot de 2007-12-04 17:43 -------
I wonder what goes wrong with building cpp at stage 3(?). :)
ana build # for f in $(find ./ -type f -name cpp); do ls -l $f; PATH=$(dirname
$f) $f -dM /dev/null | grep __DBL_MAX__; done
-rwxr-xr-x 1 root root 279196 4. Dez 18:20 ./gcc/cpp
#define __DBL_MAX__ ½.3200111555710001e+310
-rwxr-xr-x 1 root root 258066 4. Dez 17:47 ./stage1-gcc/cpp
#define __DBL_MAX__ 1.7976931348623157e+308
-rwxr-xr-x 1 root root 389702 4. Dez 18:09 ./prev-gcc/cpp
#define __DBL_MAX__ 1.7976931348623157e+308
LDFLAGS="-Wl,-O1 -Wl,--hash-style=gnu -Wl,--as-needed -Wl,--sort-common"
STAGE1_CFLAGS="-O2"
BOOT_CFLAGS=" -O2 -mtune=nocona -march=nocona -pipe"
(Hint for bad defines from:
http://forums.gentoo.org/viewtopic-p-4569014.html#4569014)
--
h dot mth at web dot de changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |h dot mth at web dot de
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (8 preceding siblings ...)
2007-12-04 17:43 ` h dot mth at web dot de
@ 2007-12-05 13:01 ` h dot mth at web dot de
2008-01-21 20:41 ` stefaan dot deroeck at gmail dot com
` (30 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: h dot mth at web dot de @ 2007-12-05 13:01 UTC (permalink / raw)
To: gcc-bugs
------- Comment #10 from h dot mth at web dot de 2007-12-05 13:01 -------
Replacing -O2 with -O1 in BOOT_CFLAGS hides the error here. At least one can
install gcc 4.3 again. :)
Gentoo users would edit toolchain.eclass:
--- /usr/portage/eclass/toolchain.eclass.orig 2007-12-05 13:58:52.000000000
+0100
+++ /usr/portage/eclass/toolchain.eclass 2007-12-05 13:58:59.000000000
+0100
@@ -1467,7 +1467,7 @@
# In general gcc does not like optimization, and add -O2 where
# it is safe. This is especially true for gcc 3.3 + 3.4
- replace-flags -O? -O2
+ replace-flags -O? -O1
# ... sure, why not?
strip-unsupported-flags
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (9 preceding siblings ...)
2007-12-05 13:01 ` h dot mth at web dot de
@ 2008-01-21 20:41 ` stefaan dot deroeck at gmail dot com
2008-02-04 20:31 ` dirtyepic at gentoo dot org
` (29 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: stefaan dot deroeck at gmail dot com @ 2008-01-21 20:41 UTC (permalink / raw)
To: gcc-bugs
------- Comment #11 from stefaan dot deroeck at gmail dot com 2008-01-21 20:18 -------
I can reproduce with gcc 4.3 snapshot 20080118
--
stefaan dot deroeck at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |stefaan dot deroeck at gmail
| |dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (10 preceding siblings ...)
2008-01-21 20:41 ` stefaan dot deroeck at gmail dot com
@ 2008-02-04 20:31 ` dirtyepic at gentoo dot org
2008-02-06 1:20 ` bangerth at dealii dot org
` (28 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: dirtyepic at gentoo dot org @ 2008-02-04 20:31 UTC (permalink / raw)
To: gcc-bugs
------- Comment #12 from dirtyepic at gentoo dot org 2008-02-04 20:30 -------
no one cares that make profiledbootstrap on 64bit targets miscompiles the
preprocessor when BOOT_CFLAGS="-O2" ?
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] Building libstdc++-v3: include/limits: stray '\275' in program
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (11 preceding siblings ...)
2008-02-04 20:31 ` dirtyepic at gentoo dot org
@ 2008-02-06 1:20 ` bangerth at dealii dot org
2008-02-06 9:28 ` [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken rguenth at gcc dot gnu dot org
` (27 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: bangerth at dealii dot org @ 2008-02-06 1:20 UTC (permalink / raw)
To: gcc-bugs
------- Comment #13 from bangerth at dealii dot org 2008-02-06 01:19 -------
Re-confirmed here:
http://gcc.gnu.org/ml/gcc/2008-02/msg00066.html
--
bangerth at dealii dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |bangerth at dealii dot org
Status|UNCONFIRMED |NEW
Ever Confirmed|0 |1
Last reconfirmed|0000-00-00 00:00:00 |2008-02-06 01:19:36
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (12 preceding siblings ...)
2008-02-06 1:20 ` bangerth at dealii dot org
@ 2008-02-06 9:28 ` rguenth at gcc dot gnu dot org
2008-02-06 10:00 ` rguenth at gcc dot gnu dot org
` (26 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-06 9:28 UTC (permalink / raw)
To: gcc-bugs
------- Comment #14 from rguenth at gcc dot gnu dot org 2008-02-06 09:28 -------
Err well. Adjusted the summary to something more descriptive, marked as
a regression (4.2 did profiledbootstrap ok).
--
rguenth at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |rguenth at gcc dot gnu dot
| |org
GCC target triplet| |x86_64-*-*
Keywords| |build, wrong-code
Known to work| |4.2.2
Summary|Building libstdc++-v3: |[4.3 Regression]
|include/limits: stray '\275'|profiledbootstrap is broken
|in program |
Target Milestone|--- |4.3.0
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (13 preceding siblings ...)
2008-02-06 9:28 ` [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken rguenth at gcc dot gnu dot org
@ 2008-02-06 10:00 ` rguenth at gcc dot gnu dot org
2008-02-06 11:20 ` rguenth at gcc dot gnu dot org
` (25 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-06 10:00 UTC (permalink / raw)
To: gcc-bugs
------- Comment #15 from rguenth at gcc dot gnu dot org 2008-02-06 09:59 -------
profiledbootstrap on x86_64 fails for me with
checking for x86_64-unknown-linux-gnu-gcc... /space/rguenther/obj/./gcc/xgcc
-B/space/rguenther/obj/./gcc/ -B/usr/local/x86_64-unknown-linux-gnu/bin/
-B/usr/local/x86_64-unknown-linux-gnu/lib/ -isystem
/usr/local/x86_64-unknown-linux-gnu/include -isystem
/usr/local/x86_64-unknown-linux-gnu/sys-include
checking for suffix of object files... configure: error: cannot compute suffix
of object files: cannot compile
See `config.log' for more details.
make[2]: *** [configure-stagefeedback-target-libgcc] Error 1
make[2]: Leaving directory `/space/rguenther/obj'
make[1]: *** [stagefeedback-bubble] Error 2
make[1]: Leaving directory `/space/rguenther/obj'
make: *** [profiledbootstrap] Error 2
obj> cat stage_current
stagefeedback
obj> tail x86_64-unknown-linux-gnu/libgcc/config.log
configure:2588: /space/rguenther/obj/./gcc/xgcc -B/space/rguenther/obj/./gcc/
-B/usr/local/x86_64-unknown-linux-gnu/bin/
-B/usr/local/x86_64-unknown-linux-gnu/lib/ -isystem
/usr/local/x86_64-unknown-linux-gnu/include -isystem
/usr/local/x86_64-unknown-linux-gnu/sys-include -c -g -O2 -fprofile-use
conftest.c >&5
<built-in>:0: internal compiler error: in real_to_decimal, at real.c:1656
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
configure:2591: $? = 1
where this ICE also occurs for empty conftest.c and if only preprocessing
using the stagefeedback cc1.
backtrace for just ./cc1 -quiet -E /dev/null
#0 fancy_abort (file=0xb530b2 "../../trunk/gcc/real.c", line=1656,
function=0xb53b20 "real_to_decimal") at ../../trunk/gcc/diagnostic.c:659
#1 0x000000000065d352 in real_to_decimal (str=0x7fff736bd500
"��s�\177",
r_orig=<value optimized out>, buf_size=<value optimized out>, digits=17,
crop_trailing_zeros=0) at ../../trunk/gcc/real.c:1656
#2 0x0000000000471275 in builtin_define_with_hex_fp_value (
macro=0x7fff736bd620 "__DBL_MAX__", type=<value optimized out>, digits=1,
hex_str=<value optimized out>, fp_suffix=0xb0b55b "",
fp_cast=0xb29315 "%s") at ../../trunk/gcc/c-cppbuiltin.c:859
#3 0x00000000004714c1 in builtin_define_float_constants (
name_prefix=0xaff950 "DBL", fp_suffix=0xb0b55b "", fp_cast=0xb29315 "%s",
type=0x2b19379cf780) at ../../trunk/gcc/c-cppbuiltin.c:205
#4 0x0000000000471ab3 in c_cpp_builtins (pfile=0xecbf70)
at ../../trunk/gcc/c-cppbuiltin.c:516
#5 0x0000000000467c9a in finish_options () at ../../trunk/gcc/c-opts.c:1508
#6 0x0000000000468069 in c_common_init () at ../../trunk/gcc/c-opts.c:1243
#7 0x000000000047341e in c_objc_common_init ()
at ../../trunk/gcc/c-objc-common.c:72
#8 0x00000000006d0410 in toplev_main (argc=0, argv=<value optimized out>)
at ../../trunk/gcc/toplev.c:2130
#9 0x00002b19377a4154 in __libc_start_main () from /lib64/libc.so.6
#10 0x0000000000403f49 in _start ()
in frame 3, the hex float looks good:
(gdb) print buf
$3 = "0x0.", 'f' <repeats 13 times>, "8p1024\000
so it must be indeed the real_to_decimal monster that is miscompiled with
-O2 -fprofile-use (given the 'correct' profile).
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (14 preceding siblings ...)
2008-02-06 10:00 ` rguenth at gcc dot gnu dot org
@ 2008-02-06 11:20 ` rguenth at gcc dot gnu dot org
2008-02-06 20:51 ` rguenth at gcc dot gnu dot org
` (24 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-06 11:20 UTC (permalink / raw)
To: gcc-bugs
------- Comment #16 from rguenth at gcc dot gnu dot org 2008-02-06 11:19 -------
Just re-building real.o with removing the real.gc* files before fixes the bug,
so to trigger this bug you indeed need the profile-feedback.
-O2 -fbranch-probabilities -finline-functions is enough to trigger the failure,
-fno-tree-vrp -fno-tree-dominator-opts (that is, disabling jump threading)
fixes
it.
It looks like rtd_divmod goes wrong, as it returns 141 for
(gdb) print *num
$90 = {cl = 1, decimal = 0, sign = 0, signalling = 0, canonical = 0,
uexp = 1024, sig = {0, 0, 18446744073709549568}}
(gdb) print *den
$91 = {cl = 1, decimal = 0, sign = 0, signalling = 0, canonical = 0,
uexp = 1017, sig = {3455452434845676643, 2207960133917416517,
16708042751414900190}}
called from
/* At this point, PTEN should contain the nearest power of 10 smaller
than R, such that this division produces the first digit.
Using a divide-step primitive that returns the complete integral
remainder avoids the rounding error that would be produced if
we were to use do_divide here and then simply multiply by 10 for
each subsequent digit. */
digit = rtd_divmod (&r, &pten);
though pten also looks funny here.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (15 preceding siblings ...)
2008-02-06 11:20 ` rguenth at gcc dot gnu dot org
@ 2008-02-06 20:51 ` rguenth at gcc dot gnu dot org
2008-02-07 20:48 ` ubizjak at gmail dot com
` (23 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-06 20:51 UTC (permalink / raw)
To: gcc-bugs
------- Comment #17 from rguenth at gcc dot gnu dot org 2008-02-06 20:51 -------
P2 - this should not block the release (it's not that profiledbootstrap was
never
broken in released compilers). It's also hard to analyze (no, I'm not on it,
volunteers welcome).
--
rguenth at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Priority|P3 |P2
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (16 preceding siblings ...)
2008-02-06 20:51 ` rguenth at gcc dot gnu dot org
@ 2008-02-07 20:48 ` ubizjak at gmail dot com
2008-02-08 14:22 ` ubizjak at gmail dot com
` (22 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-07 20:48 UTC (permalink / raw)
To: gcc-bugs
------- Comment #18 from ubizjak at gmail dot com 2008-02-07 20:47 -------
(In reply to comment #17)
> P2 - this should not block the release (it's not that profiledbootstrap was
> never
> broken in released compilers). It's also hard to analyze (no, I'm not on it,
> volunteers welcome).
It looks that ten_to_ptwo() returns wrong result for argumens >= 7.
Try this debug session:
< non-profiled (correct) cc1>:
(gdb) set args -E -quiet /dev/null
(gdb) break real_to_decimal
Breakpoint 3 at 0x599580: file ../../gcc-svn/trunk/gcc/real.c, line 1453.
(gdb) run
Breakpoint ...
(gdb) p *ten_to_ptwo (7)
$1 = {cl = 1, decimal = 0, sign = 0, signalling = 0, canonical = 0,
uexp = 426, sig = {159020156881263929, 14298703881791668535,
10644899600020376799}}
---
<profiled (wrong) cc1>:
<repeat all of the above commands>
(gdb) p *ten_to_ptwo (7)
$1 = {cl = 1, decimal = 0, sign = 0, signalling = 0, canonical = 0,
uexp = 423, sig = {5021769561911101971, 862341187440057448,
11372220470965069550}}
You can see that the results are different.
Suprisingly, result when passing 6 as an argument differs only for one LSB.
(correct):
(gdb) p *ten_to_ptwo (6)
$2 = {cl = 1, decimal = 0, sign = 0, signalling = 0, canonical = 0,
uexp = 213, sig = {5834422113351499776, 4377335499248575995,
14012984643248170709}}
(wrong):
(gdb) p *ten_to_ptwo (6)
$2 = {cl = 1, decimal = 0, sign = 0, signalling = 0, canonical = 0,
uexp = 213, sig = {5834422113351499776, 4377335499248575995,
14012984643248170708}}
Luckily, ten_to_ptwo() from real.c has only ~20 lines.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (17 preceding siblings ...)
2008-02-07 20:48 ` ubizjak at gmail dot com
@ 2008-02-08 14:22 ` ubizjak at gmail dot com
2008-02-08 20:16 ` ubizjak at gmail dot com
` (21 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-08 14:22 UTC (permalink / raw)
To: gcc-bugs
------- Comment #19 from ubizjak at gmail dot com 2008-02-08 14:21 -------
The core of the problem is, that for profiled bootstrap, the call to
normalize() from do_multiply() is simply gone. Gone in the sense, that
normalize() is neither inlined at the call site, neither is called in
do_multiply(). This leads to various observed strange effects when real numbers
are processed.
for (k = j; k < SIGSZ * 2; k += 2)
{
unsigned long bi = b->sig[k / 2];
if (k & 1)
bi >>= HOST_BITS_PER_LONG / 2;
else
bi &= ((unsigned long)1 << (HOST_BITS_PER_LONG / 2)) - 1;
u.sig[k / 2] = ai * bi;
}
>>>>> normalize (&u);
inexact |= do_add (rr, rr, &u, 0);
}
}
Adding __attirbute__((noinline)) to protptype of normalize() is enough to fix
profiled bootstrap:
Index: real.c
===================================================================
--- real.c (revision 132182)
+++ real.c (working copy)
@@ -98,7 +98,7 @@
static void clear_significand_below (REAL_VALUE_TYPE *, unsigned int);
static bool div_significands (REAL_VALUE_TYPE *, const REAL_VALUE_TYPE *,
const REAL_VALUE_TYPE *);
-static void normalize (REAL_VALUE_TYPE *);
+static void normalize (REAL_VALUE_TYPE *) __attribute__((noinline));
static bool do_add (REAL_VALUE_TYPE *, const REAL_VALUE_TYPE *,
const REAL_VALUE_TYPE *, int);
So, what upsets gcc's inliner/profiler/whatever to drop the call?
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |ubizjak at gmail dot com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (18 preceding siblings ...)
2008-02-08 14:22 ` ubizjak at gmail dot com
@ 2008-02-08 20:16 ` ubizjak at gmail dot com
2008-02-08 20:25 ` ubizjak at gmail dot com
` (20 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-08 20:16 UTC (permalink / raw)
To: gcc-bugs
------- Comment #20 from ubizjak at gmail dot com 2008-02-08 20:15 -------
(In reply to comment #19)
> So, what upsets gcc's inliner/profiler/whatever to drop the call?
Correction, normalize() gets inlined together with lshift_significand(), but
there is something wrong with inlined version. Adding attribute ((noinline))
only to lshift_significand() doesn't fix the failure.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (19 preceding siblings ...)
2008-02-08 20:16 ` ubizjak at gmail dot com
@ 2008-02-08 20:25 ` ubizjak at gmail dot com
2008-02-08 21:18 ` ubizjak at gmail dot com
` (19 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-08 20:25 UTC (permalink / raw)
To: gcc-bugs
------- Comment #21 from ubizjak at gmail dot com 2008-02-08 20:25 -------
Following patch that forces inlining of normalize breaks normal bootstrap in
exactly the same way, so it looks that there is something wrong with inliner.
Index: real.c
===================================================================
--- real.c (revision 132182)
+++ real.c (working copy)
@@ -98,7 +98,7 @@
static void clear_significand_below (REAL_VALUE_TYPE *, unsigned int);
static bool div_significands (REAL_VALUE_TYPE *, const REAL_VALUE_TYPE *,
const REAL_VALUE_TYPE *);
-static void normalize (REAL_VALUE_TYPE *);
+static inline void normalize (REAL_VALUE_TYPE *)
__attribute__((always_inline));
static bool do_add (REAL_VALUE_TYPE *, const REAL_VALUE_TYPE *,
const REAL_VALUE_TYPE *, int);
<built-in>:0: internal compiler error: in real_to_decimal, at
real.c:1656<built-in>:0: internal compiler error: in real_to_decimal, at
real.c:1656
Please submit a full bug report,
with preprocessed source if appropriate.
See <http://gcc.gnu.org/bugs.html> for instructions.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (20 preceding siblings ...)
2008-02-08 20:25 ` ubizjak at gmail dot com
@ 2008-02-08 21:18 ` ubizjak at gmail dot com
2008-02-08 21:29 ` rguenth at gcc dot gnu dot org
` (18 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-08 21:18 UTC (permalink / raw)
To: gcc-bugs
------- Comment #22 from ubizjak at gmail dot com 2008-02-08 21:17 -------
Created an attachment (id=15119)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15119&action=view)
Executable test that shows the failure.
This is the test that shows the inliner failure:
[uros@localhost ~]$ gcc -O2 test.c
[uros@localhost ~]$ ./a.out
uexp=213
sig[0]=5834422113351499776
sig[1]=4377335499248575995
sig[2]=14012984643248170709
[uros@localhost ~]$ gcc -O2 -DINLINE test.c
[uros@localhost ~]$ ./a.out
uexp=212
sig[0]=11668844226702999552
sig[1]=1837141970856070134
sig[2]=9579225209565564328
The first result is correct, the second is not.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug bootstrap/33992] [4.3 Regression] profiledbootstrap is broken
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (21 preceding siblings ...)
2008-02-08 21:18 ` ubizjak at gmail dot com
@ 2008-02-08 21:29 ` rguenth at gcc dot gnu dot org
2008-02-09 21:05 ` [Bug middle-end/33992] [4.3 Regression] profiledbootstrap is broken (miscompiling a function with inlining) ubizjak at gmail dot com
` (17 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-08 21:29 UTC (permalink / raw)
To: gcc-bugs
------- Comment #23 from rguenth at gcc dot gnu dot org 2008-02-08 21:29 -------
Now that we have a testcase.
--
rguenth at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
Priority|P2 |P1
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug middle-end/33992] [4.3 Regression] profiledbootstrap is broken (miscompiling a function with inlining)
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (22 preceding siblings ...)
2008-02-08 21:29 ` rguenth at gcc dot gnu dot org
@ 2008-02-09 21:05 ` ubizjak at gmail dot com
2008-02-09 21:06 ` ubizjak at gmail dot com
` (16 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-09 21:05 UTC (permalink / raw)
To: gcc-bugs
------- Comment #24 from ubizjak at gmail dot com 2008-02-09 21:04 -------
Created an attachment (id=15125)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15125&action=view)
Shorter and more evil testcase
This is much shorter testcase that basically consists of two functions,
do_test() and do_add() in addition to a couple of realnum handling functions.
The test however exposes more problems.
We still have:
[uros@localhost test]$ gcc -O2 t1.c
[uros@localhost test]$ ./a.out
[uros@localhost test]$ gcc -O2 -DINLINE t1.c
[uros@localhost test]$ ./a.out
Aborted
There is new problem exposed:
As can be seen in the testcase, there is a call to normalize() in do_add()
function. Since we call do_add() with arg *a of rvc_zero and arg *b of
rvc_normal class of numbers, this call is effectively a dead code.
However, when the call to normalize() is removed, the test always fail, no
matter if normalize is forced to be inlined or not.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug middle-end/33992] [4.3 Regression] profiledbootstrap is broken (miscompiling a function with inlining)
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (23 preceding siblings ...)
2008-02-09 21:05 ` [Bug middle-end/33992] [4.3 Regression] profiledbootstrap is broken (miscompiling a function with inlining) ubizjak at gmail dot com
@ 2008-02-09 21:06 ` ubizjak at gmail dot com
2008-02-10 8:23 ` [Bug middle-end/33992] [4.3 Regression] Miscompiles a function, breaks profiledbootstrap ubizjak at gmail dot com
` (15 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-09 21:06 UTC (permalink / raw)
To: gcc-bugs
------- Comment #25 from ubizjak at gmail dot com 2008-02-09 21:06 -------
Critical, IMO.
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
Severity|normal |critical
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug middle-end/33992] [4.3 Regression] Miscompiles a function, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (24 preceding siblings ...)
2008-02-09 21:06 ` ubizjak at gmail dot com
@ 2008-02-10 8:23 ` ubizjak at gmail dot com
2008-02-10 10:45 ` [Bug middle-end/33992] [4.3 Regression] Miscompiles a function with inlining, " ubizjak at gmail dot com
` (14 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-10 8:23 UTC (permalink / raw)
To: gcc-bugs
------- Comment #26 from ubizjak at gmail dot com 2008-02-10 08:23 -------
Following is significantly minimized testcase:
--cut here--
#include <stdlib.h>
#include <stdio.h>
struct real_value
{
unsigned long sig[3];
};
#ifdef INLINE
static inline void __attribute__((always_inline))
#else
static void __attribute__((noinline))
#endif
normalize (struct real_value *r)
{
int j;
for (j = 0; ; j++)
if (r->sig[2] & ((unsigned long)1 << (63 - j)))
break;
printf ("%i\n", j);
if (j)
abort ();
}
static void __attribute__((noinline))
do_test (struct real_value *r)
{
int i;
for (i = 0; i < 2; ++i)
normalize (r);
}
int main()
{
struct real_value r;
r.sig[0] = 0ull;
r.sig[1] = 0ull;
r.sig[2] = 0x8000000000000001ull;
do_test (&r);
return 0;
}
--cut here--
[uros@localhost test]$ gcc -O2 -DINLINE t1.c
[uros@localhost test]$ ./a.out
63
Aborted
[uros@localhost test]$ gcc -O2 t1.c
[uros@localhost test]$ ./a.out
0
0
Note, that it looks like MSB is not detected in normalize().
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug middle-end/33992] [4.3 Regression] Miscompiles a function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (25 preceding siblings ...)
2008-02-10 8:23 ` [Bug middle-end/33992] [4.3 Regression] Miscompiles a function, breaks profiledbootstrap ubizjak at gmail dot com
@ 2008-02-10 10:45 ` ubizjak at gmail dot com
2008-02-10 11:11 ` [Bug middle-end/33992] [4.3 Regression] Miscompiles " ubizjak at gmail dot com
` (13 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-10 10:45 UTC (permalink / raw)
To: gcc-bugs
------- Comment #27 from ubizjak at gmail dot com 2008-02-10 10:44 -------
Minimum testcase, fails also for 32bits:
--cut here--
extern void abort ();
void __attribute__((noinline))
bar (unsigned long long i)
{
if (i)
abort ();
}
void __attribute__((always_inline))
foo (unsigned long long *r)
{
int i;
for (i = 0; ; i++)
if (*r & ((unsigned long long)1 << (63 - i)))
break;
bar (i);
}
void __attribute__((noinline))
do_test (unsigned long long *r)
{
int i;
for (i = 0; i < 2; ++i)
foo (r);
}
int main()
{
unsigned long long r = 0x8000000000000001ull;
do_test (&r);
return 0;
}
--cut here--
[uros@localhost test]$ gcc -O1 t1.c && ./a.out
[uros@localhost test]$ gcc -O2 t1.c && ./a.out
Aborted
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
GCC target triplet|x86_64-*-* |
Summary|[4.3 Regression] Miscompiles|[4.3 Regression] Miscompiles
|a function, breaks |a function with inlining,
|profiledbootstrap |breaks profiledbootstrap
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug middle-end/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (26 preceding siblings ...)
2008-02-10 10:45 ` [Bug middle-end/33992] [4.3 Regression] Miscompiles a function with inlining, " ubizjak at gmail dot com
@ 2008-02-10 11:11 ` ubizjak at gmail dot com
2008-02-10 12:28 ` ubizjak at gmail dot com
` (12 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-10 11:11 UTC (permalink / raw)
To: gcc-bugs
------- Comment #28 from ubizjak at gmail dot com 2008-02-10 11:10 -------
This testcase can be used to analyze the failure:
--cut here--
extern void abort ();
void __attribute__((noinline))
bar (unsigned long long i)
{
if (i)
abort ();
}
#ifdef INLINE
void __attribute__((always_inline))
#else
void __attribute__((noinline))
#endif
foo (unsigned long long *r)
{
int i;
for (i = 0; ; i++)
if (*r & ((unsigned long long)1 << (63 - i)))
break;
bar (i);
}
void __attribute__((noinline))
do_test (unsigned long long *r)
{
int i;
for (i = 0; i < 2; ++i)
foo (r);
}
int main()
{
unsigned long long r = 0x8000000000000001ull;
do_test (&r);
return 0;
}
--cut here--
[uros@localhost test]$ gcc -O2 t1.c && ./a.out
[uros@localhost test]$ gcc -O2 -DINLINE t1.c && ./a.out
Aborted
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
Summary|[4.3 Regression] Miscompiles|[4.3 Regression] Miscompiles
|a function with inlining, |function with inlining,
|breaks profiledbootstrap |breaks profiledbootstrap
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug middle-end/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (27 preceding siblings ...)
2008-02-10 11:11 ` [Bug middle-end/33992] [4.3 Regression] Miscompiles " ubizjak at gmail dot com
@ 2008-02-10 12:28 ` ubizjak at gmail dot com
2008-02-10 12:31 ` [Bug tree-optimization/33992] " ubizjak at gmail dot com
` (11 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-10 12:28 UTC (permalink / raw)
To: gcc-bugs
------- Comment #29 from ubizjak at gmail dot com 2008-02-10 12:27 -------
The problem is in forwprop4 pass:
When testcase from #28 is compiled with "-O2 -DINLINE", forwprop4 pass
transforms:
--cut here--
;; Function do_test (do_test)
do_test (r)
{
long long unsigned int shifttmp.49;
long long unsigned int D.1576;
...
<bb 6>:
# i_19 = PHI <0(2), i_4(5)>
D.1576_13 = *r_3(D);
shifttmp.49_16 = D.1576_13 & 0x08000000000000000;
if (shifttmp.49_16 != 0)
goto <bb 5>;
else
goto <bb 3>;
...
}
--cut here--
to
--cut here--
;; Function do_test (do_test)
Replaced 'shifttmp.49_16 != 0' with '0' (-> !!!!!!!)
do_test (r)
{
long long unsigned int D.1576;
...
<bb 6>:
# i_19 = PHI <0(2), i_4(5)>
D.1576_13 = *r_3(D);
goto <bb 3>;
...
}
--cut here--
The replacement is wrong.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (28 preceding siblings ...)
2008-02-10 12:28 ` ubizjak at gmail dot com
@ 2008-02-10 12:31 ` ubizjak at gmail dot com
2008-02-10 13:46 ` ubizjak at gmail dot com
` (10 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-10 12:31 UTC (permalink / raw)
To: gcc-bugs
------- Comment #30 from ubizjak at gmail dot com 2008-02-10 12:31 -------
tree-optimization
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
Component|middle-end |tree-optimization
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (29 preceding siblings ...)
2008-02-10 12:31 ` [Bug tree-optimization/33992] " ubizjak at gmail dot com
@ 2008-02-10 13:46 ` ubizjak at gmail dot com
2008-02-10 13:49 ` pinskia at gcc dot gnu dot org
` (9 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-10 13:46 UTC (permalink / raw)
To: gcc-bugs
------- Comment #31 from ubizjak at gmail dot com 2008-02-10 13:45 -------
Patch in testing:
Index: tree-ssa-forwprop.c
===================================================================
--- tree-ssa-forwprop.c (revision 132202)
+++ tree-ssa-forwprop.c (working copy)
@@ -376,7 +376,7 @@ forward_propagate_into_cond (tree cond_e
tree op1 = TREE_OPERAND (cond, 1);
rhs0 = GIMPLE_STMT_OPERAND (def_stmt, 1);
tmp = combine_cond_expr_cond (TREE_CODE (cond), boolean_type_node,
- fold_convert (TREE_TYPE (op1), rhs0),
+ fold_convert (TREE_TYPE (rhs0),
rhs0),
op1, !single_use0_p);
}
/* If that wasn't successful, try the second operand. */
@@ -393,7 +393,7 @@ forward_propagate_into_cond (tree cond_e
rhs1 = GIMPLE_STMT_OPERAND (def_stmt, 1);
tmp = combine_cond_expr_cond (TREE_CODE (cond), boolean_type_node,
op0,
- fold_convert (TREE_TYPE (op0), rhs1),
+ fold_convert (TREE_TYPE (rhs1),
rhs1),
!single_use1_p);
}
/* If that wasn't successful either, try both operands. */
@@ -402,7 +402,7 @@ forward_propagate_into_cond (tree cond_e
&& rhs1 != NULL_TREE)
tmp = combine_cond_expr_cond (TREE_CODE (cond), boolean_type_node,
rhs0,
- fold_convert (TREE_TYPE (rhs0), rhs1),
+ fold_convert (TREE_TYPE (rhs1), rhs1),
!(single_use0_p && single_use1_p));
}
else if (TREE_CODE (cond) == SSA_NAME)
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
AssignedTo|unassigned at gcc dot gnu |ubizjak at gmail dot com
|dot org |
Status|NEW |ASSIGNED
Last reconfirmed|2008-02-06 01:19:36 |2008-02-10 13:45:55
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (30 preceding siblings ...)
2008-02-10 13:46 ` ubizjak at gmail dot com
@ 2008-02-10 13:49 ` pinskia at gcc dot gnu dot org
2008-02-10 14:32 ` ubizjak at gmail dot com
` (8 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: pinskia at gcc dot gnu dot org @ 2008-02-10 13:49 UTC (permalink / raw)
To: gcc-bugs
------- Comment #32 from pinskia at gcc dot gnu dot org 2008-02-10 13:49 -------
- fold_convert (TREE_TYPE (op1), rhs0),
+ fold_convert (TREE_TYPE (rhs0),
rhs0),
op1, !single_use0_p);
Wait, you don't need the fold_convert here then. Since rhs0 is already be
TREE_TYPE (rhs0) :).
Seriously this needs more thinking that patch. Since after your patch, we have
a type mismatch.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (31 preceding siblings ...)
2008-02-10 13:49 ` pinskia at gcc dot gnu dot org
@ 2008-02-10 14:32 ` ubizjak at gmail dot com
2008-02-10 14:33 ` ubizjak at gmail dot com
` (7 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-10 14:32 UTC (permalink / raw)
To: gcc-bugs
------- Comment #33 from ubizjak at gmail dot com 2008-02-10 14:31 -------
(In reply to comment #32)
> Seriously this needs more thinking that patch. Since after your patch, we have
> a type mismatch.
I will leave this problem to tree expert then.
(BTW: There is another fold_convert in forward_propagate_comparison(),
tree-ssa-forwprop.c, line 788, that IMO has type mismatch.
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
AssignedTo|ubizjak at gmail dot com |unassigned at gcc dot gnu
| |dot org
Status|ASSIGNED |NEW
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (32 preceding siblings ...)
2008-02-10 14:32 ` ubizjak at gmail dot com
@ 2008-02-10 14:33 ` ubizjak at gmail dot com
2008-02-10 22:09 ` rguenth at gcc dot gnu dot org
` (6 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-10 14:33 UTC (permalink / raw)
To: gcc-bugs
------- Comment #34 from ubizjak at gmail dot com 2008-02-10 14:32 -------
BTW: Although patch from Comment #31 is probably not correct, it fixes
profiledbootstrap.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (33 preceding siblings ...)
2008-02-10 14:33 ` ubizjak at gmail dot com
@ 2008-02-10 22:09 ` rguenth at gcc dot gnu dot org
2008-02-10 22:37 ` rguenth at gcc dot gnu dot org
` (5 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-10 22:09 UTC (permalink / raw)
To: gcc-bugs
------- Comment #35 from rguenth at gcc dot gnu dot org 2008-02-10 22:08 -------
The source is very likely a type mismatch in the IL (the foldings are correct,
but also strictly unneccessary). In principle it goes as follows. We have
a comparison
op0 OP op1
where op0 and op1 are required to be trivially convertible to each other. Now
we look up the definition of any of op0 or op1, where their RHS are required
to be trivially convertible to their LHS and so the RHSes are trivially
convertible to each other. Thus we don't need the conversion - for the
final IL that is - but since fold likes to have more precise types we
convert either operand to the type of the other before calling fold.
So far goes the theory - of course it all breaks down if one of the above
assumption does not hold.
I will have a look.
--
rguenth at gcc dot gnu dot org changed:
What |Removed |Added
----------------------------------------------------------------------------
AssignedTo|unassigned at gcc dot gnu |rguenth at gcc dot gnu dot
|dot org |org
Status|NEW |ASSIGNED
Last reconfirmed|2008-02-10 13:45:55 |2008-02-10 22:08:45
date| |
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (34 preceding siblings ...)
2008-02-10 22:09 ` rguenth at gcc dot gnu dot org
@ 2008-02-10 22:37 ` rguenth at gcc dot gnu dot org
2008-02-10 22:44 ` rguenth at gcc dot gnu dot org
` (4 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-10 22:37 UTC (permalink / raw)
To: gcc-bugs
------- Comment #36 from rguenth at gcc dot gnu dot org 2008-02-10 22:36 -------
Actually the wrong types are caused by loop IM which replaces the bittest
D.1210_17 = D.1208_13 >> 63;
D.1211_2 = (int) D.1210_17;
if (D.1211_1 != 0)
by
shifttmp.48_12 = 0x8000000000000000;
shifttmp.48_16 = shifttmp.48_12 & D.1208_13;
if (shifttmp.48_16 != 0)
but forgets to update the type of the zero. Testing the patch.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (35 preceding siblings ...)
2008-02-10 22:37 ` rguenth at gcc dot gnu dot org
@ 2008-02-10 22:44 ` rguenth at gcc dot gnu dot org
2008-02-11 1:35 ` dirtyepic at gentoo dot org
` (3 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-10 22:44 UTC (permalink / raw)
To: gcc-bugs
------- Comment #37 from rguenth at gcc dot gnu dot org 2008-02-10 22:43 -------
Created an attachment (id=15129)
--> (http://gcc.gnu.org/bugzilla/attachment.cgi?id=15129&action=view)
patch
I'm bootstrapping and regtesting it (it fixes the testcase). Maybe someone
can do a profiledbootstrap with it as well (otherwise I will do tomorrow).
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (36 preceding siblings ...)
2008-02-10 22:44 ` rguenth at gcc dot gnu dot org
@ 2008-02-11 1:35 ` dirtyepic at gentoo dot org
2008-02-11 8:28 ` rguenth at gcc dot gnu dot org
` (2 subsequent siblings)
40 siblings, 0 replies; 42+ messages in thread
From: dirtyepic at gentoo dot org @ 2008-02-11 1:35 UTC (permalink / raw)
To: gcc-bugs
------- Comment #38 from dirtyepic at gentoo dot org 2008-02-11 01:34 -------
yep, profiledbootstrap w/ BOOT_CFLAGS="-O2" on x86_64 is working now. i'll try
to get it tested on ppc64 to see if it fixes PR34720 too.
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (37 preceding siblings ...)
2008-02-11 1:35 ` dirtyepic at gentoo dot org
@ 2008-02-11 8:28 ` rguenth at gcc dot gnu dot org
2008-02-11 9:00 ` ubizjak at gmail dot com
2008-02-15 13:40 ` ubizjak at gmail dot com
40 siblings, 0 replies; 42+ messages in thread
From: rguenth at gcc dot gnu dot org @ 2008-02-11 8:28 UTC (permalink / raw)
To: gcc-bugs
------- Comment #39 from rguenth at gcc dot gnu dot org 2008-02-11 08:27 -------
Subject: Bug 33992
Author: rguenth
Date: Mon Feb 11 08:27:00 2008
New Revision: 132234
URL: http://gcc.gnu.org/viewcvs?root=gcc&view=rev&rev=132234
Log:
2008-02-11 Uros Bizjak <ubizjak@gmail.com>
Richard Guenther <rguenther@suse.de>
PR tree-optimization/33992
* tree-ssa-loop-im.c (rewrite_bittest): Fixup the type of
the zero we compare against.
* gcc.c-torture/execute/pr33992.c: New testcase.
Added:
trunk/gcc/testsuite/gcc.c-torture/execute/pr33992.c
Modified:
trunk/gcc/ChangeLog
trunk/gcc/testsuite/ChangeLog
trunk/gcc/tree-ssa-loop-im.c
--
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (38 preceding siblings ...)
2008-02-11 8:28 ` rguenth at gcc dot gnu dot org
@ 2008-02-11 9:00 ` ubizjak at gmail dot com
2008-02-15 13:40 ` ubizjak at gmail dot com
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-11 9:00 UTC (permalink / raw)
To: gcc-bugs
------- Comment #40 from ubizjak at gmail dot com 2008-02-11 08:59 -------
Richi, thanks for the fix.
Fixed.
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
Status|ASSIGNED |RESOLVED
Resolution| |FIXED
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread* [Bug tree-optimization/33992] [4.3 Regression] Miscompiles function with inlining, breaks profiledbootstrap
2007-11-04 1:37 [Bug bootstrap/33992] New: Building libstdc++-v3: include/limits: stray ‘\275’ in program lindevel at gmx dot net
` (39 preceding siblings ...)
2008-02-11 9:00 ` ubizjak at gmail dot com
@ 2008-02-15 13:40 ` ubizjak at gmail dot com
40 siblings, 0 replies; 42+ messages in thread
From: ubizjak at gmail dot com @ 2008-02-15 13:40 UTC (permalink / raw)
To: gcc-bugs
------- Comment #41 from ubizjak at gmail dot com 2008-02-15 13:39 -------
*** Bug 34720 has been marked as a duplicate of this bug. ***
--
ubizjak at gmail dot com changed:
What |Removed |Added
----------------------------------------------------------------------------
OtherBugsDependingO|34720 |
nThis| |
CC| |tgall dot foo at gmail dot
| |com
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=33992
^ permalink raw reply [flat|nested] 42+ messages in thread