Re: Identifying source file locations.

["Zack Weinberg" <zackw at stanford dot edu>]

On Sun, Jun 10, 2001 at 01:14:16AM -0400, Daniel Berlin wrote:

...
> So all this work is pretty much useless except for error messages,
> until we cover the case of doing something in the RTL.
> No offense or anything.
> It's just a large undertaking to go and implement, and it won't buy us
> anything without going all the way and doing something about
> NOTE_LINE_NUMBER.

Improving how well we locate the problem in error messages is not a
useless goal.  Also, all this work will have to be done in order to
deal with NOTE_LINE_NUMBER properly.  It doesn't help if every insn is
carefully labelled with its source location, if the front ends are
providing garbage to label them with.

I am nervous about adding additional information to every insn.  It
will push INSN and BARRIER rtxes over the line into the next bigger GC
bucket (JUMP_INSN, CALL_INSN, and CODE_LABEL insns are already there).
Of course, there are plenty of ways to reclaim the wasted memory,
starting with "more non-power-of-two GC buckets!"

-- 
zw                I met a traveller from an antique land
                  Who said: 'Your name is Ozymandias,
                  king of kings, and I claim my five pounds!'
                  	-- Del Cotter