any string of numbers given, such as (1.1), yeld zero. such as; Python 2.2.2 (#1, Mar 26 2003, 15:04:51) [GCC 3.2.2] on linux2 Type "help", "copyright", "credits" or "license" for more information. >>> int(1.1) 0 >>> int(5.3) 0 >>> int(3.5) 0 there was a similar problem with perl and the pentium 4, it was resolved by installing 5.8. i have recompiled with many different flags to no avail. Reproducible: Always Steps to Reproduce: get a system with a p4, try the function. Actual Results: returned a zero Expected Results: here's the output from another computer; >>> int(1.1) 1 >>> int(5.3) 5 >>> int(3.5) 3 i'm running a pentium4 2.53ghz/533
it's come to my attention that the problem is derived from a corruption of glibc. Using the -march=pentium4 flag while compiling glibc creates this issue.
alastair, there's also a TONNE of ebuilds in there -- please remove the unnecessary ones.
seemant, dev-lang/python isn't beefcake anymore. openfriday, please paste emerge info have you tried compiling glibc with a lower optimisation? i remember there was a warning in /etc/make.conf about using -march=pentium4, but it doesn't seem to be there anymore. but excessive optimisation in with -march=pentium4 is not recommended.
this problem can be fixed by recompiling glibc with lower optimizations, i would recommend implementing something in the glibc ebuild to disallow the use of -march=pentium4. I have recompiled it with -march=pentium3 and this issue has been resolved.
sorry folks, but this isn't a python bug but a glibc/over-optimisation bug. there some people in the forum with the same problem. could you please re-assign to someone who maintains glibc? thanks :) http://forums.gentoo.org/viewtopic.php?t=43373 we strip flags most from the CFLAGS but we don't strip -march and -mcpu.
As the new policy is that no stripping of flags should be done, but a list of known good ones should be made available, I reassign so that -march=pentium4 can be added to the "known bad list".
This bug results from a broken glibc modf(). See bug 14069.
Known glibc/gcc bug. This is handled already.
*** Bug 20082 has been marked as a duplicate of this bug. ***