On 30.09.23 14:53, Torbjörn Granlund wrote:
> Stefan Liebler <s...@linux.ibm.com> writes:
>
> The returned limb was retrieved from v6 instead of v2.
> This is also observable in failing testcase t-fat.
>
> While I agree the code looks suspicious, I fail to trigger any test
> suite failure for any Z/arch configuration.
>
> How exactly did you trigger it?
>
> (I try to avoid fixing any GMP bugs without making sure we trigger it
> with our testsuite and test system setup.)
>
>
I'm using a z16 with Fedora 37 (gcc 12.2.1). There I've just run:
<gmp-sources>$ ./.bootstrap
<gmp-build-dir>$ <gmp-sources>/configure
<gmp-build-dir>$ make
<gmp-build-dir>$ make check
../../../gmp/test-driver: line 112: 2635694 Aborted
(core dumped) "$@" >> "$log_file" 2>&1
FAIL: t-fat
(gdb) bt
#0 __pthread_kill_implementation (threadid=<optimized out>,
signo=<optimized out>, no_tid=no_tid@entry=0) at pthread_kill.c:44
#1 0x000003fff7ca1466 in __pthread_kill_internal (signo=<optimized
out>, threadid=<optimized out>) at pthread_kill.c:78
#2 0x000003fff7c51640 in __GI_raise (sig=sig@entry=6) at
../sysdeps/posix/raise.c:26
#3 0x000003fff7c345c0 in __GI_abort () at abort.c:79
#4 0x000003fff7e8de08 in __gmp_assert_fail
(filename=filename@entry=0x100c01e
"/home/stli/libgmpDir/gmp/tests/mpn/t-fat.c", linenum=linenum@entry=138,
expr=expr@entry=0x100c058 "r == 0")
at /home/stli/libgmpDir/gmp/assert.c:57
#5 0x00000000010030ca in check_functions () at
/home/stli/libgmpDir/gmp/tests/mpn/t-fat.c:138
#6 0x0000000001002a04 in main () at
/home/stli/libgmpDir/gmp/tests/mpn/t-fat.c:305
│ 133 for (i = 0; i < 2; i++)
│ 134 {
│ 135 xp[0] = 1296;
│ 136 r = mpn_divexact_by3c (wp, xp, (mp_size_t) 1,
CNST_LIMB(0));
│ 137 ASSERT_ALWAYS (wp[0] == 432);
│ > 138 ASSERT_ALWAYS (r == 0);
│ 139 }
In my case, before calling mpn_divexact_by3c v6 was {0x1, 0x0} and thus
mpn_bdiv_dbm1c returns 0x1, which is ANDed with 3 and then also returned
by mpn_divexact_by3c. Therefore the test fails by chance.
I assume, you are using a different gcc version.
_______________________________________________
gmp-devel mailing list
gmp-devel@gmplib.org
https://gmplib.org/mailman/listinfo/gmp-devel
