On Thu, 1 Jun 2006, Norman Gray wrote:
> On 2006 Jun 1 , at 13.28, Peter W. Draper wrote:
> >
> > I've tried to reproduce this without success, so I'm guessing this
> > is some local difficulty, say a bad sed (truncating) or corrupt
> > file somewhere?
>
> What've you done? You've mentally beamed electric rays to my poor
> machine and now it's working! And I didn't touch a thing, honest!
> This is a Good Thing (albeit slightly disturbing).
Not the first time I've peered over someone's shoulder and suddenly all is
well! But doesn't really solve your problem.
> > On Wed, 31 May 2006, Norman Gray wrote:
> >
> >> I've just done a from-the-top checkout, bootstrap and configure. The
> >> latter failed with
> >>
> >> ...
> >> checking fpu_control.h presence... yes
> >> checking for fpu_control.h... yes
> >> checking where to find Tcl/Tk 8.4+... unknown
> >> checking for star2html... (cached) star2html
> >> updating cache ../../config.cache
> >> configure: error: conditional "HAVE_ITCL" was never defined.
> >> Usually this means the macro was only invoked conditionally.
> >> configure: error: /bin/sh './configure' failed for ccdpack
> >> configure: error: /bin/sh './configure' failed for applications
>
> > In the past we've dealt with such issues by using the presence of
> > various files to check the sanity of the build order (the
> > old .INSTALLED_xxx, and .BUILT files). Maybe an idea we can
> > resurrect, although it might be less useful without stages that
> > ever removes them (no ./mk deinstall to remove .INSTALLED_xxx).
>
> I tried a tweak to ccdpack's configure.ac:
>
> ptolemy:ccdpack> cvs diff configure.ac
> Index: configure.ac
> ===================================================================
> RCS file: /cvs/applications/ccdpack/configure.ac,v
> retrieving revision 1.53
> diff -u -r1.53 configure.ac
> --- configure.ac 28 Feb 2006 21:05:04 -0000 1.53
> +++ configure.ac 1 Jun 2006 14:58:40 -0000
> @@ -59,7 +59,6 @@
> AC_MSG_RESULT(no)
> fi
> LIBS="$my_save_LIBS"
> - AM_CONDITIONAL(HAVE_ITCL, [test -n "$have_itcl"])
> have_blt=1
> my_save_LIBS="$LIBS"
> @@ -79,6 +78,11 @@
> LDFLAGS=$my_save_LDFLAGS
> fi
> +if test -z "$have_itcl"; then
> + echo "*** I can't find itcl -- I can't proceed" >&2
> + echo "*** (have you done 'make configure-deps'?)" >&2
> + exit 1
> +fi
> # Dependencies must be on one line
> STAR_DECLARE_DEPENDENCIES([build], [ard ast blt chr cnf fio generic
> grp hds idi ifd itcl ndf ndg one par pcs pgplot prm psx tcl tk trn])
> ptolemy:ccdpack>
>
> The HAVE_ITCL AM conditional wasn't being used anywhere that I could
> find, so either something subtle's happening or it's become
> redundant. At any rate, ccdpack builds OK from an otherwise fresh
> checkout, with this change.
>
> I added a manual test that it had in fact found itcl. It's not very
> sophisticated, but it's testing the right thing and appears to work.
> Shall I commit this?
May as well, but I think I've messed this up. You shouldn't use
AM_CONDITIONAL in branched code, plus as you say the Makefile.am doesn't
use it anyway. It used to control the lifting of binaries from ITCL, but
that's no longer used, so they should be removed. I'll have a further look
at this after you've committed the patch.
BTW, I was suggesting that we test if "make configure-deps" has been run
before running the top-level configure, rather than tuning the tests that
depend on this. Wouldn't a AC_CONFIG_SRCDIR() for some product of "make
configure-deps" stop that from running?
Cheers,
Peter.
|