Re: Fix libstdc++/39629 (partial) + Fix operator precedence in _ShiftMin1

[Paolo Carlini <paolo dot carlini at oracle dot com>]

Hi,
> I wonder if the last fails might be test case fails having to do with
> comparison of literals.  I'll try to do some casting with
> uint_fast64_t if you haven't already.
I will look into the 4 remaining fails as soon as possible but at moment
I have no idea what is going wrong. Any help would be appreciated.
> Should the templates force usage of long long integers?  In the
> standard?  It seems bothersome if a 32-bit system has to supply such a
> seemingly strange type to use.
No idea. I do not understand what do you mean by "forcing" or seemingly
strange type. The types at issues are perfectly normal stdint.h types,
already used elsewhere in <random>, in the standard. I didn't follow
some late details of <random> but I know people from Fermilab and other
labs, extremely knowledgeable about these issues were involved, I trust
those people. Really, I think we should just be be careful that longs
are 64-bit or 32-bit depending on the target. More correctly, if we want
to be *sure* about the size of a type, we want a stdint.h type (but
then, until PR 448 gets closed, we *cannot* assume stdint.h to be
available everywhere, this is another story, hopefully soon obsolete)

Paolo.