Re: demand_empty_rest_of_line and ignore_rest_of_line

[Hans-Peter Nilsson <hp at bitrange dot com>]

On Sat, 24 Apr 2004, Andreas Schwab wrote:
> Hans-Peter Nilsson <hp@bitrange.com> writes:
> > On Sat, 24 Apr 2004, Andreas Schwab wrote:
> >> CRIS doesn't as of current gcc mainline, and probably never did.
> > I assume you mean s/gcc/glibc/?.
>
> No.  gcc for cris does not (re-)define TARGET_ASM_FILE_START_APP_OFF, so
> the assembler never goes into NO_APP mode.

That conclusion is incorrect.  #NO_APP for cris-* is output
through default_file_start.  It seems there are multiple blessed
ways to get that #NO_APP out.  Hmm, I thought that kind of
multiplicity was something Zack W disliked but apparently not. ;-)

> > No, at least for a while it (m68k-linux) didn't.  See for
> > example the gcc-3.3 branch as of Mon Feb 23 18:58:32 GMT 2004.
>
> I can't find any relevant changes in this time frame under config/m68k on
> the 3.3 branch.

I would have expected some #include removals.  You want to look
at config.gcc too.  Maybe the it was all in the ASM_FILE_START
revamp anyway.  I compiled and checked but only guessed the
reason.

>  IIRC m68k-linux never redefined ASM_FILE_START and still
> inherits TARGET_ASM_FILE_START_APP_OFF from m68k.h.

Well, compile a file and see for yourself, given the source
dates I stated.  Note I mentioned the 3.3 branch, where there's
no TARGET_ASM_FILE_START_APP_OFF.

> Maybe the #APP/#NO_APP switching should be removed and input scrubbing
> enabled all the time, given how few targets actually disable it.

See binutils archives from last time this came up (search for
"no_app").  It saves as much as 1% off the compile time.  I
think *more* targets should use it, but most would need to tweak
their md:s to avoid redundant spaces.

brgds, H-P