[<prev] [next>] [<thread-prev] [day] [month] [year] [list]
Date: Fri, 14 Dec 2012 14:55:02 -0500
From: Rich Felker <dalias@...ifal.cx>
To: musl@...ts.openwall.com
Subject: Re: spandsp testsuite, feenableexcept() is missing
On Fri, Dec 14, 2012 at 11:49:26PM +0400, ojab wrote:
> On 14.12.2012 22:58, Rich Felker wrote:
> >On Fri, Dec 14, 2012 at 06:30:28PM +0400, ojab wrote:
> >>>Hi list,
> >>>
> >>>spandsp testsuite build fails with
> >>>> >/root/spandsp/tests/v17_tests.c:311: undefined reference to `feenableexcept'
> >>>
> >>>AFAIU all functions from fenv.h are not implemented. Is there plans
> >>>to implement them before 1.0/at all?
> >fenv.h is implemented. The issue is that feenableexcept is not part of
> >the standard for fenv.h; it's an extension added by glibc for turning
> >on trapping exceptions. This usage is non-portable (as far as I know,
> >many machines don't support trapping exceptions) and I've never seen a
> >program actually making use of them. It looks like spandsp is just
> >using this function in its test suite, probably to test some
> >exception-relted behavior. I think this should be fixed upstream,
> >since it won't work on other non-GNU systems either.
> >
>
> Ouch, stupid me, I thought that I've checked posix about it.
> Sorry for the noise.
No problem. It's still useful to have a record that at least some
program was trying to use it, even if only for testing purposes; it
helps evaluate whether it would ever make sense to add it. However, if
you're going to support trapping exceptions, there are a few fenv
functions which become slightly more complicated or more expensive,
and as long as nothing actually wants trapping exceptions, I think it
makes the most sense just to say "they're not supported and we offer
no interface to turn them on"; then, the rest of fenv can safely
assume they're off.
Rich
Powered by blists - more mailing lists
Confused about mailing lists and their use?
Read about mailing lists on Wikipedia
and check out these
guidelines on proper formatting of your messages.
