On 2007-12-29 23:25:39 +0100, Vincent Lefevre wrote: > * Some functions were failing when some global flag was set before > the call, and unfortunately most tests were done with all flags > cleared. Now, in the generic tests, all the global flags are > set before a test with a probability 1/2 (part of r5115).
Unfortunately, there was a bug in these tests, with the consequence that the value wasn't checked (so, these tests could only detect assertion failures or freeze). I've just fixed it and another bug in reciprocal trig/hyperbolic functions which wasn't detected because of that. If GCC calls these functions with all the flags cleared, it isn't affected by this bug. Moreover there is now an mpfrlint script (available via Subversion) that can be used to detect bad constructs; it should have detected the above bug, but was incomplete. Again, fixed. -- Vincent Lefèvre <[EMAIL PROTECTED]> - Web: <http://www.vinc17.org/> 100% accessible validated (X)HTML - Blog: <http://www.vinc17.org/blog/> Work: CR INRIA - computer arithmetic / Arenaire project (LIP, ENS-Lyon)
