Обсуждение: PostgreSQL 8.1.5: regression tests fail on sparc

Hi,

PostgreSQL 8.1.5 have a problem with division by zero on sparc.

Solaris 9 sparc, gcc 4.0.2, 4.1.1:

$ ./configure --enable-thread-safety --disable-nls --without-perl
--without-python --without-krb5 --without-openssl --without-readline
...
$ make && make check
...
=======================1 of 98 tests failed.
=======================

The differences that caused some tests to fail can be viewed in the
file `./regression.diffs'.  A copy of the test summary that you see
above is saved in the file `./regression.out'.

*** ./expected/errors.out       Sat Feb 12 01:15:11 2005
--- ./results/errors.out        Thu Nov 23 18:14:08 2006
***************
*** 300,308 **** select 1/0::int8; ERROR:  division by zero select 1::int2/0;
! ERROR:  division by zero select 1/0::int2;
! ERROR:  division by zero select 1::numeric/0; ERROR:  division by zero select 1/0::numeric;
--- 300,310 ---- select 1/0::int8; ERROR:  division by zero select 1::int2/0;
! ERROR:  floating-point exception
! DETAIL:  An invalid floating-point operation was signaled. This probably
meansan out-of-range result or an invalid operation, such as division by zero. select 1/0::int2;
! ERROR:  floating-point exception
! DETAIL:  An invalid floating-point operation was signaled. This probably
meansan out-of-range result or an invalid operation, such as division by zero. select 1::numeric/0; ERROR:  division by
zeroselect 1/0::numeric; 

======================================================================

Thanks a lot!

--
Alex Deiter

Дейтер Александр Валериевич <tiamat@komi.mts.ru> writes:
> PostgreSQL 8.1.5 have a problem with division by zero on sparc.
> Solaris 9 sparc, gcc 4.0.2, 4.1.1:

We've not heard that reported before, so I wonder whether there's not
something broken about your compiler or libc.

The code in question is in int2div() in utils/adt/int.c:
   if (arg2 == 0)       ereport(ERROR,               (errcode(ERRCODE_DIVISION_BY_ZERO),
errmsg("divisionby zero"))); 
   result = arg1 / arg2;

It's a bit hard to see how a compiler could get this wrong, but maybe
it's thinking it can evaluate the division before calling errfinish?
        regards, tom lane

Дейтер Александр Валериевич wrote:
> Hi,
>
> PostgreSQL 8.1.5 have a problem with division by zero on sparc.
>
> Solaris 9 sparc, gcc 4.0.2, 4.1.1:
>
> $ ./configure --enable-thread-safety --disable-nls --without-perl
> --without-python --without-krb5 --without-openssl --without-readline
> ...
>

I found that you logged this as gcc bug:
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=29968

However, SunStudio is much better for SPARC platform and now is free.
GCC does not have good code optimalizer for SPARC.

        Zdenek