COFF weak symbols

COFF weak symbols

[Mark E.]

Greetings,

I'm interested have added or in adding an extension to COFF (or at 
least DJGPP's COFF) to have weak symbols. I've already added 
the .weak directive to the Oct 1st snapshot, but I wanted to check 
and see how you wanted me proceed (or even if I should 
proceed). I'm not an object file expert, but I know just enough to 
know I want weak symbols. The linker already seems to support 
weak symbols, it would just be a matter of deciding where in the 
symbol record to store the weak symbol bit (though I have an idea).

I tried e-mailing Ian but he seems to be on vacation.

Mark

-- 
Mark E. snowball3@usa.net
http://members.xoom.com/snowball3/

Re: COFF weak symbols

[Ian Lance Taylor]

   From: "Mark E." <snowball3@usa.net>
   Date: Sat, 3 Oct 1998 11:25:18 +0000

   I'm interested have added or in adding an extension to COFF (or at 
   least DJGPP's COFF) to have weak symbols. I've already added 
   the .weak directive to the Oct 1st snapshot, but I wanted to check 
   and see how you wanted me proceed (or even if I should 
   proceed). I'm not an object file expert, but I know just enough to 
   know I want weak symbols. The linker already seems to support 
   weak symbols, it would just be a matter of deciding where in the 
   symbol record to store the weak symbol bit (though I have an idea).

   I tried e-mailing Ian but he seems to be on vacation.

You only sent me the mail Thursday evening.  I can certainly
sympathize with how it feels when you don't get an immediate reply to
your mail--I've been there myself--but the reality is that dealing
with binutils issues from the net is an unpaid activity for me, so
there is no guaranteed response time.  As I expect you've already
seen, I replied to your message earlier this morning.

Of course, unless you need my personal input, which for an issue like
this you don't, it's probably better to send such a message to the
mailing list in the first.

The main issue with any extension like this is how to store the
information in the object file.  Once that has been determined, you
need to enhance BFD to transfer the information between the generic
internal format and the object file external format.

For COFF, I would recommend defining a new storage class C_WEAK as a
parallel to C_EXT and C_STAT.  It should be moderately straightforward
to add BFD, assembler, and linker support for this.

Ian

Re: COFF weak symbols

[Mark E.]

>    From: "Mark E." <snowball3@usa.net>
>    Date: Sat, 3 Oct 1998 11:25:18 +0000
> 
>    I tried e-mailing Ian but he seems to be on vacation.
> 
> You only sent me the mail Thursday evening.  I can certainly
> sympathize with how it feels when you don't get an immediate reply to your
> mail--I've been there myself--but the reality is that dealing with

Actually, I said you were on vacation because I checked Dejanews 
and found you hadn't posted anything in a week so I made the 
rather dumb assumption (as I shouldn't have done) that were you 
were off for while. Sorry about that.

I'm on the mailing list, but I don't see any activity. I guess all the 
action is with the egcs group!

-- 
Mark E. snowball3@usa.net
http://members.xoom.com/snowball3/

Re: COFF weak symbols

[Robert Lipe]

Mark E. wrote:

> I'm interested have added or in adding an extension to COFF (or at    
> least DJGPP's COFF) to have weak symbols. I've already added the      
> .weak directive to the Oct 1st snapshot, but I wanted to check and    
> see how you wanted me proceed (or even if I should proceed). I'm not  
> an object file expert, but I know just enough to                      

I know of at least one other person that's tried it.  After discussions
with the author of the attached patch, the author of weak support in
OpenServer, and myself, I think this was abandoned.

Whether it was abandoned becuase of funkiness in the interoperability
requirement of OpenServer or just lack of motivation, I really can't
recall.   Since SCO had a free linker that did this, it may have well
been the motivation thing.

The attached patch might serve as a starting place.  It seemed like a
lot of the infrastructure was already present for other object file
formats.

I don't know the copyright status of these patches.

RJL