Re: old_procfs.h and large file environment

From: Nick Maclaren (nmm1_at_cus.cam.ac.uk)
Date: 11/29/04

  • Next message: Roger Leigh: "Re: A GNU autotools question" 
    Date: 29 Nov 2004 11:38:10 GMT

    In article <slrncqm10q.fed.dave@fly.srk.fer.hr>,
    Drazen Kacar <dave@fly.srk.fer.hr> writes:
    |>
    |> > Putting my implementor's hat on, I don't think that this example is
    |> > going to cause programs to misbehave - but I am FAR from certain
    |> > that there are no such examples, where it won't. In particular,
    |> > I know that the errno change DOES cause conforming, portable C90
    |> > programs to misbehave, with at least some valid Studio 9 options.
    |>
    |> Because it's not required to set errno on underflow any more, or you had
    |> something else in mind?

    No, underflow is an operation-level exception, not a <math.h> one,
    and isn't regarded as a critical one nowadays, anyway (but I am
    not going to start THAT debate again now).

    The <math.h> functions are no longer required to set errno on
    error, and are now permitted to set errno invalidly even if there
    is no error. As far as I know, the Solaris ones take only the
    former liberty - e.g. exp(1.0e100) may not set ERANGE and sqrt(-1.0)
    or atan(2.0) may not set EDOM.

    Regards,
    Nick Maclaren.

  • Next message: Roger Leigh: "Re: A GNU autotools question"