BUG #4190: sparc64 test suite fails

The following bug has been logged online:

Bug reference: 4190
Logged by: Dennis Gilmore
Email address: dennis(at)ausil(dot)us
PostgreSQL version: 8.3.1
Operating system: Linux
Description: sparc64 test suite fails
Details:

errors test fails. the following is the output. which to me indicates it
passed but gave different output.

*** ./expected/errors.out Sat Nov 10 09:36:44 2007
--- ./results/errors.out Fri May 23 09:18:27 2008
***************
*** 304,312 ****
select 1/0;
ERROR: division by zero
select 1::int8/0;
! ERROR: division by zero
select 1/0::int8;
! ERROR: division by zero
select 1::int2/0;
ERROR: division by zero
select 1/0::int2;
--- 304,314 ----
select 1/0;
ERROR: division by zero
select 1::int8/0;
! ERROR: floating-point exception
! DETAIL: An invalid floating-point operation was signaled. This probably
means an out-of-range result or an invalid operation, such as division by
zero.
select 1/0::int8;
! ERROR: floating-point exception
! DETAIL: An invalid floating-point operation was signaled. This probably
means an out-of-range result or an invalid operation, such as division by
zero.
select 1::int2/0;
ERROR: division by zero
select 1/0::int2;

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

"Dennis Gilmore" <dennis(at)ausil(dot)us> writes:
> errors test fails. the following is the output. which to me indicates it
> passed but gave different output.

No, it failed. The code in question looks like this:

int64 arg1 = PG_GETARG_INT64(0);
int32 arg2 = PG_GETARG_INT32(1);
int64 result;

if (arg2 == 0)
ereport(ERROR,
(errcode(ERRCODE_DIVISION_BY_ZERO),
errmsg("division by zero")));

result = arg1 / arg2;

and the output you are showing indicates that the compiler thought it
could rearrange the code so that the division was executed *despite*
the protective test in front of it.

IMHO, what you have here is a compiler bug, or at least the gcc boys are
going to need to provide a damn good excuse why it's not.

BTW, kindly revert the patch you inserted in the Fedora repository.

regards, tom lane

"Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:

> IMHO, what you have here is a compiler bug, or at least the gcc boys are
> going to need to provide a damn good excuse why it's not.

You don't, perhaps have any of -ffast-math or these two suboptions which
-ffast-math enables turned on?

`-funsafe-math-optimizations'
Allow optimizations for floating-point arithmetic that (a) assume
that arguments and results are valid and (b) may violate IEEE or
ANSI standards. When used at link-time, it may include libraries
or startup files that change the default FPU control word or other
similar optimizations.

This option should never be turned on by any `-O' option since it
can result in incorrect output for programs which depend on an
exact implementation of IEEE or ISO rules/specifications for math
functions.

The default is `-fno-unsafe-math-optimizations'.

`-fno-trapping-math'
Compile code assuming that floating-point operations cannot
generate user-visible traps. These traps include division by
zero, overflow, underflow, inexact result and invalid operation.
This option implies `-fno-signaling-nans'. Setting this option
may allow faster code if one relies on "non-stop" IEEE arithmetic,
for example.

This option should never be turned on by any `-O' option since it
can result in incorrect output for programs which depend on an
exact implementation of IEEE or ISO rules/specifications for math
functions.

The default is `-ftrapping-math'.

--
Gregory Stark
EnterpriseDB http://www.enterprisedb.com
Ask me about EnterpriseDB's Slony Replication support!

The full build log is
http://sparc.koji.fedoraproject.org/koji/getfile?taskID=56849&name=build.log

the build was using gcc-4.3 with the package as built on the rest of fedora.
it builds ok sparc32 ppc ppc64 i386 x86_64 and ia64 but fails on sparc64.

I will forward everything to the gcc guys in a bug report.

Dennis

On Friday 23 May 2008, Gregory Stark wrote:
> "Tom Lane" <tgl(at)sss(dot)pgh(dot)pa(dot)us> writes:
> > IMHO, what you have here is a compiler bug, or at least the gcc boys are
> > going to need to provide a damn good excuse why it's not.
>
> You don't, perhaps have any of -ffast-math or these two suboptions which
> -ffast-math enables turned on?
>
> `-funsafe-math-optimizations'
> Allow optimizations for floating-point arithmetic that (a) assume
> that arguments and results are valid and (b) may violate IEEE or
> ANSI standards. When used at link-time, it may include libraries
> or startup files that change the default FPU control word or other
> similar optimizations.
>
> This option should never be turned on by any `-O' option since it
> can result in incorrect output for programs which depend on an
> exact implementation of IEEE or ISO rules/specifications for math
> functions.
>
> The default is `-fno-unsafe-math-optimizations'.
>
> `-fno-trapping-math'
> Compile code assuming that floating-point operations cannot
> generate user-visible traps. These traps include division by
> zero, overflow, underflow, inexact result and invalid operation.
> This option implies `-fno-signaling-nans'. Setting this option
> may allow faster code if one relies on "non-stop" IEEE arithmetic,
> for example.
>
> This option should never be turned on by any `-O' option since it
> can result in incorrect output for programs which depend on an
> exact implementation of IEEE or ISO rules/specifications for math
> functions.
>
> The default is `-ftrapping-math'.