This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: [COMMITTED] tile: define __NO_LONG_DOUBLE_MATH
- From: Joseph Myers <joseph at codesourcery dot com>
- To: Chris Metcalf <cmetcalf at ezchip dot com>
- Cc: <libc-alpha at sourceware dot org>
- Date: Wed, 25 Nov 2015 01:09:44 +0000
- Subject: Re: [COMMITTED] tile: define __NO_LONG_DOUBLE_MATH
- Authentication-results: sourceware.org; auth=none
- References: <1448399819-2804-1-git-send-email-cmetcalf at ezchip dot com> <alpine dot DEB dot 2 dot 10 dot 1511242230080 dot 10652 at digraph dot polyomino dot org dot uk> <56550015 dot 7020109 at ezchip dot com>
On Tue, 24 Nov 2015, Chris Metcalf wrote:
> as well, so it seemed uncontroversial. And I only chose 2.12 on the theory
> that they
> should have been exported then. I don't really care what version I export
That's not how things work. The version is always the version where the
symbols were actually first exported, not where they "should" have been
exported; you never add symbols to a past release version. See exp2l in
math/Versions or fallocate64 in sysdeps/unix/sysv/linux/<various 32-bit
architectures>/Versions for examples.
> them as, since there's really no way for userspace to generate a
> reference to them,
Which is why it would be optimal not to export them at all.
> what would be cleaner here and I am happy to make the change, but it does seem
> like it should be a change for every platform without long double, not just
> tile.
Well, you don't remove symbols from old versions either. Optimally the
default should be that they aren't exported, with existing no-long-double
platforms that had the exports defining some macro in their sysdeps files
(e.g. math_private.h) to cause them to be exported (and ideally causing
them to be exported as compat symbols only, given that it's not possible
for the headers to generate references to them).
--
Joseph S. Myers
joseph@codesourcery.com