Re: Live range splitting in new allocator

[Christopher Faylor <cgf at redhat dot com>]

On Mon, Jan 29, 2001 at 11:20:40AM -0800, Linus Torvalds wrote:
>In article <20010129135420.G13475@cse.cygnus.com> you write:
>>
>>In addition to the dataflow problems, another problem that I haven't had time
>>to work on is how to inform the debugger of a variables location as it moves
>>betweent he stack and multiple registers.  The current compiler is broken in
>>this regard, but LRS makes it more visible, since things can be in multiple
>>registers for a longer time.
>
>The debugger problem comes up not infrequently as a reason for gcc doing
>the wrong thing.
>
>Would it not be reasonable at some point to say "Ok, our debug
>information is going to be bad, but let's do the RightThing(tm) anyway,
>and fix the debug information _afterwards_".
>
>Now, I'm obviously biased, as I think that the main reason for a
>debugger is to get a backtrace and not much more.  But if debugging is
>holding back good register allocation, then somebody should realize that
>register allocation is just about the most important thing a compiler
>can do, and that bad debugging infrastructure will never ever get
>improved if it never gets a _reason_ to improve. 
>
>So please at least consider breaking debugging.  It _will_ get fixed -
>too many people care about it.  But it should not get fixed by
>generating bad code. 

However, if you (the GCC developers) do decide to break the debugger
(gdb), kindly be so polite as to let the debugger people formally know
how and why in a cogent email to gdb@sources.redhat.com rather than
assuming that we'll be able to glean useful information by reading
threads in gcc@gcc.gnu.org.

cgf