From: "Charles Sandmann" <sandmann AT clio DOT rice DOT edu>
Newsgroups: comp.os.msdos.djgpp
Subject: Re: DXE
Date: Sat, 11 Aug 2001 16:37:42
Organization: Aspen Technology, Inc.
Lines: 34
Message-ID: <3b755f56.sandmann@clio.rice.edu>
References: <slrn9n5b8t DOT 238t DOT marcov AT toad DOT stack DOT nl> <3B72D840 DOT 50C82D3D AT is DOT elta DOT co DOT il> <slrn9n5tug DOT kt6 DOT marcov AT toad DOT stack DOT nl>
NNTP-Posting-Host: dcloan.hou.aspentech.com
X-Trace: selma.aspentech.com 997566250 8888 10.32.115.107 (11 Aug 2001 21:44:10 GMT)
X-Complaints-To: postmaster AT aspentech DOT com
NNTP-Posting-Date: 11 Aug 2001 21:44:10 GMT
X-NewsEditor: ED-1.5.8
To: djgpp AT delorie DOT com
DJ-Gateway: from newsgroup comp.os.msdos.djgpp
Reply-To: djgpp AT delorie DOT com

> The last thing is only a hack to avoid the real problem. I'd rather be able
> to use the runtime library from the mainprogram. 

To do this you can either:
1) link with libc (-lc) if the routine is standalone and doesn't reference
  external data.  For example, a string routine works here, but malloc()
  doesn't.
2) create a transfer vector for the "call back" routines/data and have the
  loader fill them in.

> To perform some late bindings, I though about simply making some mods to the 
> DXE format (and increasing the magic):
> - DXEGEN doesn't exit, but only warn on missing symbols. the missing symbol
> names is stored in the DXE too.
> - (not needed for late binding) Instead of one pointer to return, make room
>  for a table (one int count, count pointers)
> 
> Then, the DXE loader routine must open the COFF binary for stabs, and try to
> calc an address for the missing symbols, and use it to relocate and link the
> missing symbols (both from mainprogram + libc/rtl) with the DXE.

This requires the symbol table not be stripped from the image.  If the image
is stripped it fails.  The transfer vector approach works even with stripped
images.

> If the DXE loader can't load (e.g. because the bin is stripped) the DXE load
> will fail. Because of the change magic, one loader can be used. 
> 
> Does this all pass a theoritical sanity check? Afaik it does, at least I
>    know that I can find the symbol for an address in stabs, what about the
>    other way around.

This could theoretically work, yes.  Look at some of the other DXE/DLL tools
in v2tk for examples.