DevHeads.net

Postings by Jerry James

Yet another gargantuan mathematical software update

Hi folks,

It's about time for another "update the world" moment for some of the
mathematical software we have in Fedora. I'm going to try to clear
away a big chunk of my backlogged package work with this update, so it
will be even bigger than usual. Primary goals are to switch from
atlas or the reference blas to openblas, and to drop python 2
packages.

I plan to do all of the necessary rebuilds myself. Package
maintainers, if you would rather that I did not rebuild your package
for you, please let me know. Otherwise, I will do all of these builds
in approximately 1 week from today.

Problem with mock --forcearch=s390x

I am trying to determine why the latest version of bigloo segfaults on
s390x during the build. I launched a mock build with mock -r
fedora-rawhide-s390x --forcearch=s390x --rebuild
bigloo-4.3c-1.fc30.src.rpm, then went away for several hours.

3 days vs 7 days

Koji seems to be requiring 7 days in -testing for F29. Shouldn't it
be requiring 3 days at this point?

Request for help with antlr4

We're heading for a Fedora 29 release with a broken coq/frama-c/why3
stack. I've got fixes waiting in the wings, but they're blocked on
these antlr4 bugs:

<a href="https://bugzilla.redhat.com/show_bug.cgi?id=1596974" title="https://bugzilla.redhat.com/show_bug.cgi?id=1596974">https://bugzilla.redhat.com/show_bug.cgi?id=1596974</a>
<a href="https://bugzilla.redhat.com/show_bug.cgi?id=1599015" title="https://bugzilla.redhat.com/show_bug.cgi?id=1599015">https://bugzilla.redhat.com/show_bug.cgi?id=1599015</a>

Time is running out, and the antlr4 maintainer clearly doesn't have
time to do anything about the situation. I don't have the bandwidth
to just jump in and take over. Is there anybody who can help with the
2 new Java packages required to upgrade antlr4, and the actual upgrade
itself?

Release-monitoring github issues?

I just saw a remark on another mailing list about an updated version
of a package I maintain. Sure enough, a new version was released, but
I never got email from release-monitoring.org about it. So I started
checking, and I've got nearly 2 dozen packages with new versions
upstream, but nary an email about it.

It looks like something might be broken with monitoring github. None
of the new versions have shown up on release-monitoring.org, and they
all show errors in the logs.

OpenBLAS: link with which library?

Bugs have been filed asking packagers to build with openblas instead
of atlas. But there are multiple openblas libraries.

More math updates, with sonames bumps and license updates

I've accumulated another set of package updates for the
sagemath/Macaulay2 set of packages. I would like to build these in
Rawhide in about a week. Here are the proposed changes. Please let
me know of any objections. I believe I have all dependent packages
for the soname bumps on the list to rebuild.

arb: update to 2.14.0

clisp: rebuild due to the pari update

cocoalib: update to 0.99595

eclib: update to 20180710.

Intent to retire cryptominisat4

With cvc4 1.6 building in Rawhide right now, the last Fedora user of
cryptominisat 4.x is gone. All Fedora consumers are now on
cryptominisat 5.x, which is in the cryptominisat package.
Consequently, I intend to retire the cryptominisat4 package soon. If
somebody needs it, let me know and I will give it to you instead of
retiring it.

cvc4 license change

The license of cvc4 has been corrected from GPv3+ to "Boost and BSD
and MIT". Since that is a less restrictive license, it should cause
no issues.

Review swaps

A new version of cvc4 is available. It depends on cryptominisat 5.x
instead of 4.x. Since cvc4 was the last consumer of cryptominisat 4.x
in Fedora, this means that we can finally retire the cryptominisat4
package.

On the other hand, cvc4 1.6 has new dependencies. I get to add 4
packages in order to retire 1. Um ... yay?

I need reviews for the following.

Annobin: "causes a section type conflict with..."

This afternoon, I received email from koschei, telling me that
polymake's builds have started to fail:

<a href="https://apps.fedoraproject.org/koschei/package/polymake?collection=f29" title="https://apps.fedoraproject.org/koschei/package/polymake?collection=f29">https://apps.fedoraproject.org/koschei/package/polymake?collection=f29</a>

The failure is due to this:

/builddir/build/BUILD/polymake-3.2/include/core/polymake/internal/shared_object.h:410:9:
error: ‘void pm::shared_object<Object, TParams>::leave() [with Object
= pm::sparse2d::Table<pm::Rational, false,
(pm::sparse2d::restriction_kind)0>; TParams =
{pm::AliasHandlerTag<pm::shared_alias_handler>}]’ causes a section
type conflict with ‘const pm::perl::RegistratorQueue&
polymake::common::get_regi

Builds failing with connection refused error

I don't see anything but green on status.fedoraproject.org, but the
last two builds I have submitted have both failed on s390x only with
an error that looks like this:

Traceback (most recent call last):
File "/usr/lib/python2.7/site-packages/koji/daemon.py", line 1244, in runTask
response = (handler.run(),)
File "/usr/lib/python2.7/site-packages/koji/tasks.py", line 307, in run
return koji.util.call_with_argcheck(self.handler, self.params, self.opts)
File "/usr/lib/python2.7/site-packages/koji/util.py", line 209, in
call_with_argcheck
return func(*args, **kwargs)
File "/us

NTL 11.1.0 update

I intend to update the ntl package from 11.0.0 to 11.1.0 in Rawhide.
Upstream has a policy of bumping the soname on every release, whether that
is necessary or not, so I will also rebuild all dependent packages, namely
these:

- eclib
- flint
- giac
- latte-integrale
- linbox
- pynac
- Macaulay2
- sagemath
- Singular

Since several of these packages depend on python, I plan to wait until the
python3 update is complete before starting these builds.

Cryptominisat license change

I will build cryptominisat 5.6.3 for Rawhide momentarily. With this
version, the license changes from LGPLv2 to MIT. This version also carries
an soname bump. I will rebuild the only dependent package in Fedora,
namely stp.

python-sphinx_rtd_theme advice

I somehow wound up owning python-sphinx_rtd_theme, I think probably because
I was the first person to attempt to do a build that needed it. :-)
Anyhow, I am not at all sure that I am competent to be the maintainer. I
need some advice. There is a new upstream version available, 0.4.0. I am
looking at it right now, and my attempts at unbundling fonts from this
package seem to be unraveling.

Three font families are bundled: fontawesome, Roboto Slab, and Lato. We
have all 3 of these in Fedora, but the problem is that this package wants
the web versions of these fonts, too.

VM in a logical volume

For work reasons, I need a Fedora 28 VM. I've got a bunch of disk space
set aside to create logical volumes (LVM) for exactly this purpose, with
several existing VMs. On this occasion, I decided that I no longer needed
an older Fedora VM and decided to recycle it. I got the Fedora 28 x86_64
Workstation ISO image, set that as the boot CD in virt-manager, and fired
it up.

Updates to mathematical software

Hello everyone. Months ago, I started working on updates to a couple of
our mathematical packages. But they, in turn, required other packages to
be updated, and those updates required other packages to be updated, and
the whole thing kind of snowballed. I believe that I have finally reached
a point of closure, where I can update the whole pile and have everything
still work afterwards.

I propose to do the following updates and builds in Rawhide in about a
week.

Heads up: selinux-policy-3.14.1-25.fc28 breaks GDM

I installed the latest batch of updates for F28 tonight. Since that
included a new kernel (4.16.10-300.fc28), I rebooted. The system came up
with the GDM panic screen [1]. I rebooted into the previous kernel
thinking that something might be wrong with the new one. Same result. I
rebooted again and added enforcing=0 to the kernel boot line. That
worked. I did a full SELinux relabel immediately afterwards. Nothing
relevant changed labels.

The SELinux Alert Browser says there are no alerts.

Re: [ACTION NEEDED] Missing BuildRequires: gcc/gcc-c++

On Sun, Feb 18, 2018 at 10:09 AM, Igor Gnatenko
< ... at fedoraproject dot org> wrote:
Fixed: abc, gnofract4d, libedit, lrslib.

The abe package does not actually need a C++ compiler for building on
Linux. The configure script does check for a C++ compiler, for use
with XCode on OS X. But only a C compiler is ever invoked on non-OS X
platforms.

Koji build: cannot find spec file

This is the second of two attempts to build sympy this evening, both
of which failed the same way:

<a href="https://koji.fedoraproject.org/koji/taskinfo?taskID=25124365" title="https://koji.fedoraproject.org/koji/taskinfo?taskID=25124365">https://koji.fedoraproject.org/koji/taskinfo?taskID=25124365</a>

According to mock_output.log, the spec file can't be found. What is going on?

ABI gate: internal-only shared object

Here's something I didn't expect from the new ABI gate. Which, before
I go further, I think will be a great idea nearly all of the time. I
think avoiding unintentional ABI breaks in stable releases is a worthy
goal.

But ... I maintain a package called gap. It provides what amounts to
a scripting language for doing certain types of math. It comes with a
bunch of addon packages that provide specific mathematical
capabilities. Most of them are written solely in the scripting
language, but a few have to interface with external libraries.

Orphaning some Java packages

I have orphaned some Java packages. These two were used to support
jnormaliz, which used to be part of the normaliz package. It is now a
separate project, and I don't use it, so I am not packaging it. If
somebody would like to do so, you will want to take up these two
packages:
- balloontip (balloon tip component for Swing applications)
- javasysmon (Java system monitor)

The other two were intended to be dependencies of jReality. However,
while I was still moving these packages along, jReality upstream moved
from the free version of itext to the non-free version.

Re: [Bug 1529276] New: findbugs-contrib-7.2.0.sb is available

On Wed, Dec 27, 2017 at 5:16 AM, < ... at redhat dot com> wrote:
Why am I on the

GCL and SELinux: help requested

On Sat, Oct 7, 2017 at 9:34 AM, Jerry James < ... at gmail dot com> wrote:

Build stuck setting up buildroot

I started an F27 build this morning:

<a href="https://koji.fedoraproject.org/koji/buildinfo?buildID=979845" title="https://koji.fedoraproject.org/koji/buildinfo?buildID=979845">https://koji.fedoraproject.org/koji/buildinfo?buildID=979845</a>

Hours later, all arches but 32-bit ARM have finished. The 32-bit ARM
build appears to be stuck setting up the buildroot.

soname updates, license changes, and version updates, o my!

I would like to update a bunch of the numerical packages in the near
future. I will do the updates for Rawhide only at first, although if
everything seems stable I may update F-27 as well. Here is the
proposed list of changes. Package maintainers and co-maintainers are
CCed to this email.

arb: update from 2.10.0 to 2.11.1

cddlib: polymake used to link with both the GMP and non-GMP versions
of cddlib. We have some ugly linker tricks in the cddlib package to
support both this and other package's use of cddlib.

Symbol aliasing issue

Tonight I am trying to figure out why gfan failed the mass rebuild.
I'm a bit confused, so I'm writing in the hopes that one of you
recognizes what is going on.

The gfan library is linked with libcddgmp. Libcdd can be built with
or without gmp support. Polymake, weirdly, wants to link with both
versions, so libcdd is built without gmp support and libcddgmp is
built with gmp support. Polymake expects the gmp version to have
symbols that are suffixed with _gmp.

Help with an s390x build failure

Are there any test s390x instances available for a peon like me to use
to diagnose a build failure? Failing that, does anybody with access
to an s390x instance have time to help me? The latest bigloo release
built successfully on all arches but s390x:

<a href="https://koji.fedoraproject.org/koji/taskinfo?taskID=20740401" title="https://koji.fedoraproject.org/koji/taskinfo?taskID=20740401">https://koji.fedoraproject.org/koji/taskinfo?taskID=20740401</a>

The failure comes after the main binary has been built and linked
successfully, but then is executed to compile some Scheme files. The
binary is reporting that it was invoked on an empty input file.

Thank you,

ocaml-menhir license change

I will shortly build ocaml-menhir version 20170509. With this
version, the license of the main packages changes from QPL with
exceptions to GPLv2. Also, an earlier mistake will be corrected: the
ocaml-menhir-devel subpackage should be LGPLv2 with exceptions, not
LGPLv2+ with exceptions.

Soname bumps and license changes

Work has been going on for months now to update quite a few of our
numerical packages. It looks like we've finally ironed the last
problems out, so Rawhide builds will start shortly.