H.J. Lu's frame.c patch is not included

H.J. Lu's frame.c patch is not included

[Teemu Torma]

In egcs-971105, H.J. Lu's frame.c patch is not included, event
though ChangeLog suggests it.

Sun Oct 26 11:41:49 1997  Jason Merrill  <jason@yorick.cygnus.com>

	...
	From H.J. Lu:
	* frame.c (count_fdes, add_fdes): Skip linked once FDE entries.

The code says:
      /* Skip CIEs and linked once FDE entries.  */
      if (this_fde->CIE_delta == 0 || this_fde->pc_range == 0)
	continue;

so it is not skipping entries when pc_begin == 0.
Is it a typo or just vanished somehow?  In either case, the current
code does not work unless pc_begin check is included.

Teemu

Re: H.J. Lu's frame.c patch is not included

[Jeffrey A Law]

  In message < 199711070926.KAA27169@baht.labs.trema.com >you write:
  > The code says:
  >       /* Skip CIEs and linked once FDE entries.  */
  >       if (this_fde->CIE_delta == 0 || this_fde->pc_range == 0)
  > 	continue;
  > 
  > so it is not skipping entries when pc_begin == 0.
  > Is it a typo or just vanished somehow?  In either case, the current
  > code does not work unless pc_begin check is included.
Looks like Jason re-added it today...  Must have been a goof of
some kind.


jeff

Re: H.J. Lu's frame.c patch is not included

[Jason Merrill]

>>>>> Teemu Torma <tot@Trema.COM> writes:

> Maybe the range is calculated by the assembler/linker before the linker
> removes the symbols?

You're right, my bad.

Jason

H.J. Lu's frame.c patch is not included

[Teemu Torma]

    From:  Jason Merrill <jason@cygnus.com>
    Date:  07 Nov 1997 02:31:52 -0800

     > The code says:
    >       /* Skip CIEs and linked once FDE entries.  */
    >       if (this_fde->CIE_delta == 0 || this_fde->pc_range == 0)
    > 	continue;
    
    > so it is not skipping entries when pc_begin == 0.
    > Is it a typo or just vanished somehow?  In either case, the current
    > code does not work unless pc_begin check is included.
    
    Why not?  I expected that whenever pc_begin would be 0, pc_range would also
    be 0, since both the begin and end labels would be relocated to 0.  Is this
    not the case?
    
At least it does not seem to be the case, there are entries with
zero pc_begin non-zero pc_range.  I am using gas-970915 in this case.
Maybe the range is calculated by the assembler/linker before the linker
removes the symbols?

Teemu

Re: H.J. Lu's frame.c patch is not included

[Jason Merrill]

>>>>> Teemu Torma <tot@trema.com> writes:

> In egcs-971105, H.J. Lu's frame.c patch is not included, event
> though ChangeLog suggests it.

> Sun Oct 26 11:41:49 1997  Jason Merrill  <jason@yorick.cygnus.com>

> 	...
> 	From H.J. Lu:
> 	* frame.c (count_fdes, add_fdes): Skip linked once FDE entries.

> The code says:
>       /* Skip CIEs and linked once FDE entries.  */
>       if (this_fde->CIE_delta == 0 || this_fde->pc_range == 0)
> 	continue;

> so it is not skipping entries when pc_begin == 0.
> Is it a typo or just vanished somehow?  In either case, the current
> code does not work unless pc_begin check is included.

Why not?  I expected that whenever pc_begin would be 0, pc_range would also
be 0, since both the begin and end labels would be relocated to 0.  Is this
not the case?

Jason