[Bug libc/16965] New: tst-strtod-round fails on 32-bit sparc

["aurelien at aurel32 dot net" <sourceware-bugzilla at sourceware dot org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=16965

            Bug ID: 16965
           Summary: tst-strtod-round fails on 32-bit sparc
           Product: glibc
           Version: 2.20
            Status: NEW
          Severity: normal
          Priority: P2
         Component: libc
          Assignee: aurelien at aurel32 dot net
          Reporter: aurelien at aurel32 dot net
                CC: drepper.fsp at gmail dot com
              Host: sparc-linux-gnu
            Target: sparc-linux-gnu
             Build: sparc-linux-gnu

tst-strtod-round fails on 32-bit sparc with the following errors:

strtold (0x1p-16446) returned 0x0.000000000000000100000001p-16382 not
0x0.0000000000000001p-16382 (default rounding mode)
strtold (0x1p-16446) returned 0x0.000000000000000100000001p-16382 not
0x0.0000000000000001p-16382 (FE_DOWNWARD)
strtold (0x1p-16446) returned 0x0.000000000000000100000001p-16382 not
0x0.0000000000000001p-16382 (FE_TOWARDZERO)
strtold (0x1p-16446) returned 0x0.000000000000000100000001p-16382 not
0x0.0000000000000001p-16382 (FE_UPWARD)
strtold (-0x1p-16446) returned -0x0.000000000000000100000001p-16382 not
-0x0.0000000000000001p-16382 (default rounding mode)
strtold (-0x1p-16446) returned -0x0.000000000000000100000001p-16382 not
-0x0.0000000000000001p-16382 (FE_DOWNWARD)
strtold (-0x1p-16446) returned -0x0.000000000000000100000001p-16382 not
-0x0.0000000000000001p-16382 (FE_TOWARDZERO)
strtold (-0x1p-16446) returned -0x0.000000000000000100000001p-16382 not
-0x0.0000000000000001p-16382 (FE_UPWARD)

Digging into the code, the issue is that __mpn_rshift is called with cnt = 0,
while stdlib/rshift.c explicitly says that cnts should satisfy 0 < CNT <
BITS_PER_MP_LIMB. In this case, the code end up doing a logical shift right of
the same size than the register (32 bit there), which is undefined in the C
standard.

While the problem appears on sparc for this specific value, I believe that it
is possible to find values which do not work on architectures which only use
the lower part of the register to get the shift amount. This includes for
example mips or aarch64.

Patch will follow soon on the mailing list.

-- 
You are receiving this mail because:
You are on the CC list for the bug.