Re: cmake version problem

[tlaronde%polynum.com@localhost]

Le Fri, Jul 21, 2023 at 12:49:15PM +0200, Dr. Thomas Orgis a écrit :
> Am Fri, 21 Jul 2023 12:20:17 +0200
> schrieb tlaronde%polynum.com@localhost:
> 
> > FWIW, when compiling some packages (pkgsrc-2023Q2), the compilation
> > failed with a puzzling message about some dir or some file (like
> > ".../work/cmake-build-..."; I didn't write it down, so it's from memory
> > and likely partly incorrect) not existing.
> > 
> > It happened that the cmake version on the building node was 3.9.
> > Updating to 3.26 solved the problem.
> 
> Isn't that a problem with any dependency? I didn't have the impression
> that pkgsrc really tracks required versions of dependencies. There is
> syntax for it, sure, but I guess not every upstream source properly
> documents the required cmake version properly, even. Also ? with
> buildlink3 files, I don't see custom version dependencies at all, only
> ABI breaks being possibly handled.
> 
> Maybe put things like cmake on a shortlist to update first before
> attempting any other update build?
> 

As gdt@ proposed, at least bumping the minimal cmake version required,
so that cmake is not updated for nothing every time, seems reasonable.

And for your other point, as suggested in another thread, you are
indeed right (so it seems to me) about the various versions of the
dependencies. Some pkgsrc tools verify all the dependencies and so
it seems build the more recent version needed (if two dependencies
depend on the same package but do not require the same version and one
is more recent than the one installed that happens to satisfy the other
package), and some tools don't: one end up with a dependency
satisfied about what is currently installed, and then linking with
a dshared lib of version M; and then another depedency wants an
update, so the version of the dshared lib linked against by the previous
dependency is removed and one has to hope that the ABI is compatible.

It seems more and more that pbulk is the canonical way for building.
There is no panacea for packages with numerous dependencies and in fact
too many combinations when not requiring an exact version of everything.
-- 
        Thierry Laronde <tlaronde +AT+ polynum +dot+ com>
                     http://www.kergis.com/
                    http://kertex.kergis.com/
Key fingerprint = 0FF7 E906 FBAF FE95 FD89  250D 52B1 AE95 6006 F40C