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

From: Paolo Carlini (pcarlini_at_[hidden])
Date: 2002-10-31 11:36:15

Gabriel Dos Reis wrote:

>| I see.
>| Therefore, in other terms, Nathan's analysis (see audit trail) could be 
>| *incorrect*.
>
>Nathan's analysis about numeric_limits<double> is correct.
>
In itself, sure. What I meant is: incorrect as an explanation for the 
problem we are seeing at run time with statistic_tests.cpp.

>| performed much better and also 2.95.x, which has a work around, did so: 
>| this seemed
>| to confirm Nathan's analysis.
>
>GCC 2.95.x never had <limits>.
>
Indeed. I meant (sorry for my bad english): Boost provides a replacement 
(not really a workaround, strictly speaking) for the missing <limits> 
header of 2.95.x.

>| Now, what should we do? Do you believe boost/random or its configury 
>| machinery are
>| to blame or libstdc++-v3 as shipped with gcc-3.2 and gcc-3.3?
>
>As I said earlier, the PR should be closed -- it is not a regression
>and it is fixed on mainline.
>
I agree and I will. However, what about the different failure I'm 
seeing, later at run-time, with 3.3 and 2.95.x? Probably another Gcc PR 
should be tentatively opened.

Also, *why* at run-time statistic_tests.cpp compiled by gcc-3.2 Aborts 
so soon? I think nobody knows that, right now. Nathan apparently belives 
that something is wrong with "user", i.e. boost/random. And you?

Ciao, Paolo.