Re: fix-n-continue

Re: fix-n-continue

[Geoffrey Keating]

Mike Stump <mrs@apple.com> writes:

> On Friday, July 30, 2004, at 07:04  PM, David Edelsohn wrote:
> > 	Can you make the new target bits local to darwin.h and overlap
> > some of the SVR4 bits instead of taking up the remaining generally
> > available bits for something that is Darwin-specific?
> 
> Yes.  I didn't like the overlap idea, as the bits can't be used
> simultaneously.  I moved them all into the darwin files.
> 
> Ok?

No.  This patch has many problems, most notably lack of any
explanation as to what it does.  At the very least,

> *** ./doc/invoke.texi.~1~	Mon Jul 26 19:11:21 2004
> --- ./doc/invoke.texi	Wed Jul 28 19:02:35 2004
> *************** This is by default ON.
> *** 6930,6935 ****
> --- 6930,6951 ----
>   @opindex -gfull
>   Emit debugging information for all symbols and types.
>   
> + @item -mfix-and-continue
> + @itemx -ffix-and-continue
> + @opindex mfix-and-continue
> + @opindex ffix-and-continue
> + Generate code suitable for fast turn around development.  Needed to
> + enable gdb to dynamically load @code{.o} files into already running
> + programs.
> + 
> + @item -mindirect-data
> + @itemx -findirect-data
> + @opindex mindirect-data
> + @opindex findirect-data
> + Generate code suitable for fast turn around development.  Needed to
> + enable gdb to dynamically load @code{.o} files into already running
> + programs.
> + 
>   @item -all_load
>   @opindex all_load
>   Loads all members of static archive libraries.
> --------------

Why are there two switches with identical documentation?  Why are
these target-specific switches documented as starting with -f?

Re: fix-n-continue

[Mark Mitchell]

Mike Stump wrote:

> Here are some patches to support fast turn around debugging for darwin...

Minor point:

>+   lprefix = (((name[0] == '*' || name[0] == '&')
>+               && (name[1] == 'L' || (name[1] == '"' && name[2] == 'L')))
>+              || (name[0] == '_'
>+                  && name[1] == 'O'
>+                  && name[2] == 'B'
>+                  && name[3] == 'J'
>+                  && name[4] == 'C'
>+                  && name[5] == '_'));
>
>  
>
Should just be "strncmp (name, "_OBJC_", 6)" -- unless you're convinced 
that you get inferior code that way?

-- 
Mark Mitchell
CodeSourcery, LLC
(916) 791-8304
mark@codesourcery.com