On Wed, 23 Feb 2005, Tim Jenness wrote:
> On Tue, 22 Feb 2005, Peter W. Draper wrote:
>
> > On Tue, 22 Feb 2005, Malcolm J. Currie wrote:
> >
> > > > Module: cursa
> > > > Status: failed
> > > > Build Log: http://dev.starlink.ac.uk/build/TRU64-ALPHA/logs/cursa.1109049944.log
> > >
> > > This is failing because pgp_link_adam (defined in cursa_link_adam) isn't
> > > found. Is there synopsis of how to handle PGPLOT linking these days in
> > > the GNU-tools world?
> >
> > Don't know of such a synopsis (Norman or Tim might), but in this case it
> > seems that graphpar wasn't declared as a link dependency of pgp. I've made
> > that fix.
> >
>
> Actually, I think it turns out the problem is that, yet again, the LINK
> dependencies are not being processed correctly. Brad has just been trying
> to build figaro and he is running into similar dependency problems.
>
> Figaro depends on AGI and AGI has a link dependency on TRN and GRAPHPAR.
> When Brad tries to build figaro (by installing the manifest) TRN, REF, SGS
> and GRAPHPAR are not built and so figaro fails.
>
> AGI has:
>
> STAR_DECLARE_DEPENDENCIES(link, [ref trn sgs pgp pgplot idi hds graphpar])
>
> Have these dependencies stopped being included again? I thought we had
> fixed this. Clearly we shouldn't make figaro depend on TRN and REF when it
> doesn't use them except through AGI. Is this the source of the cursa
> indirect dependency?
My build of builds turned up this problem. FIGARO does not declare that it
also depends on the link dependencies of its build dependency libraries.
i.e. it stated:
STAR_DECLARE_DEPENDENCIES(build, [pda agi grp ndf fio idi])
when it meant:
STAR_DECLARE_DEPENDENCIES([build], [pda agi grp ndf fio idi], [link])
I've made this change and updated Makefile.dependencies so it should be OK
now.
Peter.
|