[Bug 1350257] Review Request: petsc - Portable Extensible Toolkit for Scientific Computation

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



https://bugzilla.redhat.com/show_bug.cgi?id=1350257



--- Comment #21 from Dave Love <d.love@xxxxxxxxxxxxxxx> ---
Apologies for the delay, what with being away, and having to spend a
lot more time on this than I expected.

(In reply to Antonio Trande from comment #20)
> SPEC:
> http://copr-dist-git.fedorainfracloud.org/cgit/sagitter/petsc/petsc.git/
> plain/petsc.spec?id=79c67a625efcd1f39a24705c8ce881977f2bb0b1
> 
> SRPM:
> https://copr-be.cloud.fedoraproject.org/results/sagitter/petsc/fedora-
> rawhide-x86_64/00446542-petsc/petsc-3.7.2-11.fc26.src.rpm

Unfortunately I still can't run fedora-review as I get gcc ICEs which I guess
are due to memory in the largest VM I can run sensibly.  I assume that's due
to MAKE_NP=%{?_smp_mflags}, which seems to be converted to just -j and goes
wild.

> SuperLU support works with superlu-2.5.0, see previous build:
> http://copr-fe.cloud.fedoraproject.org/coprs/sagitter/petsc/build/446367/

I assume that means 5.2

> > --known-endian=$(%{__python2} -c "import sys;print sys.byteorder") \
> 
> Is this option totally useless?

Well, configure tests that anyway, doesn't it?

I still find the spec difficult to follow with the redundancy and
conditionals for things which are constant, but I guess that's not
blocking.

Please fix the build not to use just -j ("MAKE_NP=%{?_smp_mflags}"
isn't valid).  I wasted a lot of time figuring out why it won't build
in the VM I need for fedora-review -- the build was going wild,
leading to threads being killed without useful diagnostics.

The build tries to run hostname, which isn't in the BRs on f24, at
least.  I'm not sure it's worth bothering about, though, as it's a
pain to pick up with different package and binary names in different
epel/fedora versions, and not having it doesn't stop anything working.

I think the architecture condition for openblas should match what the
openblas spec uses, i.e.
%ifarch x86_64 %{ix86} armv7hl %{power64} aarch64

For el6, use environment-modules consistently for build- and
install-time requires, rather than the installation requiring Lmod.
You'll need to define _modulesdir for el6 if you use it.

Can we avoid the need for a module anyhow?  It doesn't make a lot of
sense for a library like this.  I don't understand the environment
variable usage, but presumably the values could be hard-wired for the
packaging.

I still don't understand the SETOPT_FLAGS thing, which definitely
isn't necessary, and why flags are set separately (and differently?)
at compile- and run-time.  None of that seems to be necessary.

Delete the .html files under /usr/include.  Maybe the .html should go
into a -doc package, but I think I figured that wasn't useful before.

I don't think the /usr/share/petsc*/conf directories should be in the
main packages (or even in the -devel ones?)

There is now version 3.7.3, but let's persist with this one for the review.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
You are always notified about changes to this product and component
_______________________________________________
package-review mailing list
package-review@xxxxxxxxxxxxxxxxxxxxxxx
https://lists.fedoraproject.org/admin/lists/package-review@xxxxxxxxxxxxxxxxxxxxxxx




[Index of Archives]     [Fedora Legacy]     [Fedora Desktop]     [Fedora SELinux]     [Yosemite News]     [KDE Users]     [Fedora Tools]