Email List: Xaustin-group-lX
[All Lists]

Purpose of XSI extensions to the ar(1) utility

To: <yyyyyyyyyyyyyy@xxxxxxxxxxxxx>
Subject: Purpose of XSI extensions to the ar(1) utility
From: "Schwarz, Konrad" <yyyyyyyyyyyyyy@xxxxxxxxxxx>
Date: Mon, 20 Jun 2005 11:11:55 +0200
Thread-index: AcV0zJmHf0wSoJo3R4ONCl03rtSBtAAqRbbg
Thread-topic: Purpose of XSI extensions to the ar(1) utility
The ar(1) manual page defines several XSI shaded extensions relating to
the placement of files within an archive, notably -a, -b, -i, -m, and,
to a lesser extent, -q.

I can fathom no reason for these extensions, except by contstruing the
following model of linking from an archive:

do 
        for each object file in archive _in the order the object files
are stored in the archive_
                if the object file contains any definitions of
unresolved references
                        add the object file to the output
while any object file was added during the last pass

Historically, has this been the implemented model?  Does the model still
hold in the presence of an archive symbol table?  If so, would it make
sense to codify this, XSI shaded, in the description of c99?

I suspect that this model is not viable for dynamically shared objects
and the dynamic linker, but the facility is useful for static libraries.

Konrad Schwarz

<Prev in Thread] Current Thread [Next in Thread>
  • Purpose of XSI extensions to the ar(1) utility, Schwarz, Konrad <=