Re: [boost] Trouble with statistic_tests.cpp and gcc-3.2/3.3

From: Paolo Carlini (pcarlini_at_[hidden])
Date: 2002-10-31 12:27:06

Nathan Sidwell wrote:

> that bit of code is filling a vector with increasing values and places
> infinity at the end (without checking has_infinity). For 3.2, this
> has the effect of placing 0 a the end, and therefore breaking the
> monotonic invariant. Lower_bound is allowed to assume the iterators
> point to increasing values, and that assumption is broken here. 

So, after all, you are *right*!

I mean, everything *really* boils down to the  <limits> header for 
doubles...

Therefore, nothing may be done on the Gcc release branch, on the 
contrary, the maintainer of boost/random (probably Jens Maurer, I'm 
CC-ing him) should autoconf the header for this "feature" ;-) of 
gcc-3.0/1/2 C++ run-time library.

Something similar to what present in the replacement header for 
gcc-2.95.x missing <limits> will of course do the trick...

Thanks Nathan,
Paolo.