On 02/05/2018 11:23 PM, Bill Allombert wrote: > On Mon, Feb 05, 2018 at 06:05:50PM +0100, Tobias Hansen wrote: >> On 02/05/2018 06:03 PM, Bill Allombert wrote: >>> On Mon, Feb 05, 2018 at 05:35:12PM +0100, Tobias Hansen wrote: >>>> Source: pari >>>> Version: 2.9.4-1 >>>> Severity: normal >>>> >>>> Hi there, >>>> >>>> one of the tests of cypari2 failed on the mips architectures (at least >>>> mips and mips64el) and the problem is in pari itself (this was on >>>> mips): >> Cool. Yes it is, on all other architectures the test ran fine: >> >> https://buildd.debian.org/status/package.php?p=cypari2&suite=experimental > Some ieee754 operations give unexpected results: > > The attached program give on mipsel: > a=inf ai=2147483647 b=-inf bi=2147483647 > while > a=inf ai=-2147483648 b=-inf bi=-2147483648 > is expected (and seen on other plateforms). > > Cheers,
Where does it say that this is expected? I would say it's undefined. From the c0x standard: When a finite value of real floating type is converted to an integer type other than _Bool, the fractional part is discarded (i.e., the value is truncated toward zero). If the value of the integral part cannot be represented by the integer type, the behavior is undefined. http://c0x.coding-guidelines.com/6.3.1.4.html Best, Tobias
