Re: Lazy DECL_RTL breakage for weak alias

[Mark Mitchell <mark at codesourcery dot com>]

>>>>> "Jim" == Jim Wilson <wilson@cygnus.com> writes:

    Jim> Perhaps it has something to do with debug info.  dbxout.c

I hate this kind of argument. :-)

These macros should have defined meanings.  Then, they should be set
in the way that makes sense.  Then, the rest of the compiler should
use them.  If there's not enough information, then we should add more
bits, until there is...

Anyhow, I agree that:

  It does make a wierd kind of sense though.  An alias is defined by
  this file, and hence is not an external reference.

I think we probably should just document this configuration, and say
what it means...

Thanks for working it through.

For reference, H.J's case was:

const int __new_sys_nerr = 1;
extern __typeof (__new_sys_nerr) _new_sys_nerr __attribute__ ((alias
  ("__new_sys _nerr")));

As far as I can tell, we're doing the right thing in that case, in
terms of matching 2.95.2.

How do I run the GDB testsuite?  What is a good GDB to use for
testing?

If you tell me how to take these steps, I will run the GDB testsuite
on IA32 and see if that turns up any other problems with the DECL_RTL
changes.

Thanks,

--
Mark Mitchell                   mark@codesourcery.com
CodeSourcery, LLC               http://www.codesourcery.com