Re: PATCH: Optimize ia64 linker relaxation

[James E Wilson <wilson at specifix dot com>]

On Tue, 2006-03-21 at 09:09, H. J. Lu wrote:
> This patch optimizes ia64 linker relaxation by skipping unneeded
> relaxation passes.

It also generalizes the code to support more than the current 2
relaxation passes, which you failed to mention.  You also failed to
mention why you want this, but we have discussed this before.  You want
to add another type of IA-64 linker relaxation which may require a 3rd
relaxation pass.

This looks like a reasonable cleanup, and you do have a good reason for
wanting it, so I am willing to allow it.

There is some inconsistency in how skip_relax_pass_[01] are used.  Most
places set them to TRUE/FALSE.  However, one place checks the values
against 1.  It should be using TRUE instead for consistency.

You added a new function relax_sections, without adding an explanatory
comment before it.

You have overloaded the relax_pass field, such that it means two
different things, depending on where it is being used.  However, the
comment you added in bfdlink.h only documents one of the uses.  This
needs to document both of them.

You left alone the definition of the need_finalize_relax field in the
bfd-in2.h file, which comes from the section.c file.  However, there is
no longer any finalize relax pass, so the field name and comment don't
make any sense.  This needs to be fixed somehow.  Maybe by documenting
that the name is obsolete, or maybe by changing the name to something
more meaningful.
-- 
Jim Wilson, GNU Tools Support, http://www.specifix.com