Laurent Fousse a écrit : > Hello, > > * Sylvain Chevillard [Wed, Nov 29, 2006 at 03:14:16PM +0100]: > >> Hi, >> >> There is currently a big bug in the version of MPFI of etch. >> >> I mentionned it to Nathalie Revol who told me that currently, there >> were several branches of MPFI and she didn't seem to be optimistic >> about when the developpers would patch the bug. >> >> However, it seems that the bug is know for a while and was patched in >> one of the several branches. >> >> Since your the maintainer of the package for Debian, can you fix it ? >> That is, at least include the patched branch into etch. >> > > Could you send me this patch? I think it may too late to include the > fix for etch, we're already quite far in the release process. But it's > a bug to fix anyway. > > >> Here is a minimal example of the bug : >> > [...] > >> Once compiled, just enter [-3.2, -3.1] to the prompt >> Note that -Pi lays in the interval and sin usually vanished at Pi. >> > > I see, the output interval should contain 0. Well, a patch is welcome. > > Laurent. >
a version where the bug is corrected is available at http://gforge.inria.fr/frs/download.php/468/mpfi-1.3.4-RC3.tar.gz According to C. Lauter, this version could have another problem (infinite loop on certain inputs) but he does not remember such an example. Anyway the bug is less critical than a false result. Sylvain.
