This is the mail archive of the
binutils@sources.redhat.com
mailing list for the binutils project.
Re: demand_empty_rest_of_line and ignore_rest_of_line
- From: Hans-Peter Nilsson <hp at bitrange dot com>
- To: Andreas Schwab <schwab at suse dot de>
- Cc: Nathan Sidwell <nathan at codesourcery dot com>, Ian Lance Taylor <ian at wasabisystems dot com>, binutils at sources dot redhat dot com
- Date: Sat, 24 Apr 2004 12:07:48 -0400 (EDT)
- Subject: Re: demand_empty_rest_of_line and ignore_rest_of_line
- References: <40587E7B.10705@codesourcery.com> <m3lllzo1vt.fsf@gossamer.airs.com><405979E9.9030805@codesourcery.com> <m3n052qw2g.fsf@whitebox.m5r.de>
On Fri, 23 Apr 2004, Andreas Schwab wrote:
> This means that the unknown character is no longer ignored, despite the
> comment. As a side effect a line starting with a line comment character
> not followed by APP in NO_APP mode now triggers an error instead of just a
> warning, breaking builds of glibc on m68k-linux.
For the record, there aren't supposed to *be* any comments in
#NO_APP mode; just strict formatted assembly code and #APP of
course. That's what it means.
If the glibc __sec_comment kludge is the problem, let me mention
that different patches have been sent to glibc in order to make
its kludge work properly(!) at least one that adds #NO_APP and
I'm sure would work, but it seems none have been applied.
Of course, we could do something that lets glibc emit its
.gnu.warning.* sections without assembler warnings, and avoid
using the __sec_comment kludge where the new thing is supported.
brgds, H-P