[Python-Dev] RE: Possible bug (was Re: numpy, overflow, inf, ieee, and rich comparison)

[Tim Peters]

[Charles G Waldman]
> I just repeated this test, on the following 4 platforms:
>
> 1)   SunOS fsui02 5.6 Generic_105181-21 sun4u sparc SUNW,Ultra-Enterprise
>      cc: WorkShop Compilers 4.2 30 Oct 1996 C 4.2
>
> 2)   OSF1 V4.0 878 alpha
>      DEC C V5.6-079 on Digital UNIX V4.0 (Rev. 878)
>
> 3)   IRIX64 6.5 07151432 IP27
>      MIPSpro Compilers: Version 7.2.1
>
> 4)   Linux 2.2.16 #31 SMP Thu Jun 29 14:51:26 CDT 2000 i686
>      gcc version 2.95.2 19991024 (release)
>
> On all of the above systems, math.exp(-999) returns 0.0, math.sqrt(-1)
> raises an exception, and test_math.py completes successfully.

Excellent!  Note that test_math.py was augmented to verify that:

1. exp(-huge) returns 0 without exception.
2. exp(+huge) raises OverflowError.
3. sqrt(-1) raises ValueError.

None of those were reliable across platforms before, and I'm afraid there's
strong reason to suspect one or more still won't work right under Mandrake
Linux when Python is compiled with -O (this based on the output of a C
program someone posted to c.l.py yesterday).    But it's a world better than
it used to be.

> Testing-ly (not to be confused with testily) yr's,

You bet -- everyone leave testily to me on this one.  I'm bitter enough
about floating-point in stinking C for everyone combined <0.9 wink>.