This is the mail archive of the
libc-alpha@sourceware.org
mailing list for the glibc project.
Re: Fix finite-math-only lgamma functions signgam setting (bug 19211)
- From: Joseph Myers <joseph at codesourcery dot com>
- To: Andreas Schwab <schwab at linux-m68k dot org>
- Cc: <libc-alpha at sourceware dot org>
- Date: Mon, 14 Mar 2016 22:55:30 +0000
- Subject: Re: Fix finite-math-only lgamma functions signgam setting (bug 19211)
- Authentication-results: sourceware.org; auth=none
- References: <alpine dot DEB dot 2 dot 10 dot 1511051858550 dot 26133 at digraph dot polyomino dot org dot uk> <87egbcog42 dot fsf at linux-m68k dot org> <alpine dot DEB dot 2 dot 10 dot 1603142226000 dot 30856 at digraph dot polyomino dot org dot uk> <87a8m0oes5 dot fsf at linux-m68k dot org>
On Mon, 14 Mar 2016, Andreas Schwab wrote:
> > Presumably some architecture-specific fix is required for whatever the
> > issue is with building with that combination of macros
>
> No, you cannot just #undef _LIBC.
Well, it can be fragile, but it's already done in e.g.
stdio-common/scanf17.c, another test that tries to avoid internal header
contents. I believe we have consensus that it would be desirable for most
tests not to get _LIBC defined, or include internal headers, at all. And
for -std=c99 builds it can be necessary to undefine such macros. If it
breaks, we identify the specific cause and add other _LIBC / _ISOMAC
conditionals, or undefine more macros, as needed.
--
Joseph S. Myers
joseph@codesourcery.com