This is the mail archive of the
mailing list for the libc-ports project.
Re: [RFC][PATCH v2] MIPS: IEEE 754-2008 NaN encoding support
- From: "Joseph S. Myers" <joseph at codesourcery dot com>
- To: "Maciej W. Rozycki" <macro at codesourcery dot com>
- Cc: <libc-ports at sourceware dot org>, <libc-alpha at sourceware dot org>, Doug Gilmore <Doug dot Gilmore at imgtec dot com>
- Date: Tue, 10 Sep 2013 23:24:36 +0000
- Subject: Re: [RFC][PATCH v2] MIPS: IEEE 754-2008 NaN encoding support
- Authentication-results: sourceware.org; auth=none
- References: <alpine dot DEB dot 1 dot 10 dot 1308222343480 dot 8514 at tp dot orcam dot me dot uk> <Pine dot LNX dot 4 dot 64 dot 1308231511370 dot 22185 at digraph dot polyomino dot org dot uk> <alpine dot DEB dot 1 dot 10 dot 1309032201260 dot 29360 at tp dot orcam dot me dot uk> <Pine dot LNX dot 4 dot 64 dot 1309061622410 dot 8532 at digraph dot polyomino dot org dot uk> <alpine dot DEB dot 1 dot 10 dot 1309101029530 dot 29360 at tp dot orcam dot me dot uk>
On Tue, 10 Sep 2013, Maciej W. Rozycki wrote:
> I infer you'd prefer to have this change committed even though the
> required kernel part isn't there yet, right? I have therefore set the
> minimum version required artificially to 10.0.0 so that this code is not
> accidentally activated on a kernel that doesn't support the 2008-NaN
> feature and put a FIXME note next to it. Please let me know if this is
> what you had in mind.
Yes, using 10.0.0 here seems reasonable.
More generally, that would also seem appropriate if a new architecture
port were submitted and approved before the kernel port is upstream.
Rather than require a strict ordering of components going upstream
(binutils, GCC, Linux kernel, glibc; GDB anywhere after binutils), lots of
unnecessary delays can be avoided by each component reviewing
independently, possibly in parallel, with the understanding that
incompatible changes may still be possible if a component you depend on
that's not yet upstream has such changes. (Because of circular
dependencies, you can't fully avoid that anyway; GCC for various
architectures includes information about signal frames produced by the
Linux kernel, for example.)
At the point when the kernel support goes upstream, the version in glibc
can then be changed to reflect the correct upstream kernel version (in the
case of new architecture ports, it's possible more __ASSUME_* macros may
also be defined, or conditionals removed, based on what was in the kernel
version that went upstream).
Joseph S. Myers