Re: HEAD: Kill two showstoppers at once

[Gary V. Vaughan]

On 30 Aug 2005, at 17:51, Ralf Wildenhues wrote:
> Hi Gary,

Hawdy!

> * Gary V. Vaughan wrote on Tue, Aug 30, 2005 at 06:31:42PM CEST:
> > Ralf Wildenhues wrote:
> > > Just a slight cleanup of your suggestion.  Changes: change
> > > m4_fatal to m4_warn, because we want to be forward compatible,
> >
> > Agreed.
>
> I also had to add a few newlines, else we'd end up with
>
> | # Set options
> | enable_dlopen=yesenable_win32_dll=yes
>
> see updated patch at end.

Oops.  Thanks!

> > > But there is another problem with respect to macro ordering: The
> > > settings set by the LT_INIT options appear too late in the configure
> > > script.  If I set AC_LIBTOOL_DLOPEN, everything is OK; but if I use
> > > LT_INIT([dlopen]), it appears after a macro which might override
> > > enable_dlopen.
> >
> > Which macro is that?
>
> LT_SYS_DLOPEN_SELF, but basically all the hundreds of require'ments of
> _LT_SETUP are expanded before LT_INIT as well..

Ah, okay.  We just need to move the _LT_SET_OPTIONS up the expansion  
stack.
If it didn't take arguments we could just AC_REQUIRE a wrapper, but  
it does
so we're stuck with tweaking the call in LT_INIT to:

  _LT_SHELL_INIT([_LT_SET_OPTIONS([$1])])

Should do the trick (untested!).

> > The fix belongs there I think.
>
> Are you sure?

Absolutely! ;-)

> The _LT_SETUP macro is instantiated after _LT_SET_OPTIONS in LT_INIT,
> but all its requirements are expanded before LT_INIT; see
> http://lists.gnu.org/archive/html/autoconf/2005-08/msg00100.html
> for a nice explanation about AC_REQUIRE.  :)
>
> (I wouldn't have guessed I could tell the M4 maintainer something
> about macro expansion.. ;-)

Have you seen the implementation of m4_require?  I'm thinking of doing
a C version for m4-2.0... the autoconf version is way hairy! ;-)

> > Does moving _LT_SET_OPTIONS([$1])dnl up to being the first line of
> > LT_INIT work?
>
> No, why should it?

It shouldn't. Just checking that I wasn't missing something in the  
provide
chain.

> > > Is this bug description comprehensible?  (I guess a couple of  
> > > m4_require
> > > and/or AC_REQUIRE are needed, I'm sure you'll know the correct  
> > > answer
> > > right away.)
> >
> > Maybe.  Or perhaps LT_INIT needs splitting, or _LT_SET_OPTIONS  
> > running
> > in a diversion... If you tell me the macro that is getting emitted
> > before the _LT_SET_OPTIONS expansion I should be able to figure it  
> > out.
>
> Have fun.  :)

With m4?  How could it be otherwise? :-b

> 2005-08-30  Gary V. Vaughan <address@hidden>,
>             Ralf Wildenhues <address@hidden>
>
>         * libltdl/m4/ltoptions.m4 (_LT_SET_OPTIONS): Move option macro
>         dipatch..
>         (_LT_SET_OPTION): here.  Do not fail but warn on unknown
>         options.
>         (_LT_UNLESS_OPTIONS(win32-dll), LT_OPTION_DEFINE(dlopen)):
>         Add whitespace.
>         Reported by Robert Ögren <address@hidden> and
>         Akim Demaille <address@hidden>.

Cheers,
    Gary.
--
Gary V. Vaughan      ())_.  gary@ 
{lilith.warpmail.net,gnu.org},address@hidden
Research Scientist   ( '/   http://www.tkd.kicks-ass.net
GNU Hacker           / )=   http://www.gnu.org/software/{libtool,m4}
Technical Author   `(_~)_   http://sources.redhat.com/autobook

PGP.sig
Description: This is a digitally signed message part