-no-undefined broken for HEAD

[Stephan Kulow]

On Tuesday 13 August 2002 20:21, Allan Sandfeld Jensen wrote:
> On Tuesday 13 August 2002 19:35, Stephan Kulow wrote:
> > On Tuesday 13 August 2002 17:14, Zack Rusin wrote:
> >
> > The GNU ld never checked that with libtool AFAIK. But am_edit creates
> > .closure files when it sees -no-undefined in LDFLAGS. If this is broken,
> > then it's an am_edit problem. But undefined references shoud be detected.
>
> That makes sense. So while libtool does nothing, we test ourselves if there
> is closure by doing a test compilation and creates the .closure file as a
> succes flag. But if we do the test ourselves and we expect libtool to
> ignore the option. Why do we add it to LDFLAGS in the first place?
We don't expect libtool to ignore it. It just does so. 

> Is there a reason why we cant move --no-undefined to another line, that
> am_edit can trigger on to generate the needed code?
Yes, because it belongs in LDFLAGS and is supposed to make the linker
check too

Greetings, Stephan