Re: speed up large library linking

[Howard Chu]

Alexandre Oliva wrote:
> On May 10, 2005, Ralf Wildenhues <address@hidden> wrote:
>
>
> > * Ralf Wildenhues wrote on Mon, May 09, 2005 at 10:58:01PM CEST:
> >
> > > - rewrite piecewise old archive linking
>
>
> > While reading the GNU ld manpage:
> > | A number of modifiers (mod) may immediately follow the p keyletter,  to
> > | specify variations on an operation's behavior:
> > | [...]
> > | S   Do not generate an archive symbol table.  This can speed up  build-
> > |     ing  a  large  library in several steps.  The resulting archive can
> > |     not be used with the linker.  In order to build a symbol table, you
> > |     must  omit  the S modifier on the last execution of ar, or you must
> > |     run ranlib on the archive.
>
>
> > Is there any reason we don't use this?  Should I hack up a patch for it?
>
>
> I don't think so, and Yes, please!, respectively :-)
>
> While at that, how about losing ranlib when ar already does its job?
> Ian Lance Taylor suggested we could test for this by creating an
> archive with ar, no ranlib, and trying to link an executable that
> pulls symbols from the archive.  We should probably make sure the
> symbol we bring in form the archive isn't available if the archive
> isn't linked in (e.g., from libc), just to be paranoid.
Can't you detect this just by whether "ar rs" is a valid command? If 
that fails, then that should mean ar doesn't generate the symbol table, 
and if it succeeds, then it does.

--
  -- Howard Chu
  Chief Architect, Symas Corp.       Director, Highland Sun
  http://www.symas.com               http://highlandsun.com/hyc
  Symas: Premier OpenSource Development and Support