More instantiation problems under hpux

Loren James Rittle rittle@latour.rsch.comm.mot.com
Tue Jan 15 18:47:00 GMT 2002


> I know that floating point exceptions are not enabled by default.  Could
> this be an issue here?  It might be possible to create an unnormalized
> value larger than that in the header.

Without a little bit of data, namely, type T, limits related to T,
value t and value i for the call in the test case that asserts, I
wouldn't be able to really speculate further.  I would guess `yes'.

Assume I only know enough to be dangerous: If floating point
exceptions were enabled, would it be impossible to create unnormalized
values larger than the limits expressed in the system limits header?
Or, are these two different issues?

On my platform, I don't actually know whether floating point
exceptions are enabled or disabled by default.  However, I do know
that the standard C library conversion routines (e.g. atof) can return
correct values well outside the limits encoded in the system limits
header.

I had assumed your system and sparc-sun-solaris* failed at the same
assertion for the same reason but perhaps it is only a coincidence.

Regards,
Loren



More information about the Libstdc++ mailing list