Malcolm,
On 2004 Aug 19 , at 17.12, Malcolm J. Currie wrote:
>> Nothing I can think of! ./bootstrap in the top level should make sure
>> that a file install-sh is present in that top-level directory.
>> Towards
>> the end of the bootstrap, it copies install-sh from buildsupport to
>> '.'
>
> There were many lines
>
> configure.ac: installing `./install-sh'
They come from the invocations of autoreconf in subdirectories.
Autoreconf is invoked at the top level, but because there's no
Makefile.am there, it presumes it doesn't need to create the link to
install-sh. Hence I create that link `by hand', by copying it from
buildsupport/install-sh, so that the recursive ./configure can find it
in this location.
>> Do you have the file install-sh in that directory in fact?
>
> Yes, in the cvs top-level directory.
This is weird. In the top-level configure script on my machine, right
after the stanza including the string `checking whether ln -s works' is
the very straightforward check for install-sh in '.', './..' and
'./../..', so if it's there, I'm baffled at it claiming it isn't. And
I don't think it would have an opportunity to add it after you ran the
./configure.
Does it still fail in the same way if you run it again?
> I have to admit to not being an avid reader of ./bootstrap output to
> know if there's any abnormal messages.
Neither am I. And unfortunately it's not as if errors stand out
particularly clearly in it.
> There weren't any warning messages about versions of autoconf etc.
> being
> missing. I know because I've tried ./bootstrap again, and those
> appeared.
Those are normal, as you know. I've never been completely clear which
generated logic it is which is complaining about them.
Norman
--
----------------------------------------------------------------------
Norman Gray : Physics & Astronomy, Glasgow University, UK
http://www.astro.gla.ac.uk/users/norman/ : www.starlink.ac.uk
|