Monday, July 20, 2015

VMware Workstation 11 and kwin - hangs and hiccups (solved?)

Since updating to VMware Workstation 11 (from the Gentoo vmware overlay), I've experienced a lot of hangs of my KDE environment whenever a virtual machine was running. Basically my system became unusable, which is bad if your workflow depends on accessing both Linux and (gasp!) Windows 7 (as guest). I first suspected a dbus timeout (doing the "stopwatch test" for 25s waits), but it seems according to some reports that this might be caused by buggy behavior in kwin (4.11.21). Sadly I haven't been able to pinpoint a specific bug report.

Now, I'm not sure if the problem is really 100% fixed, but at least now the lags are much smaller- and here's how to do it (kudos to matthewls and vrenn): 
  • Add to /etc/xorg.conf in the Device section
    Option "TripleBuffer" "True"
  • Create a file in /etc/profile.d with content
    __GL_YIELD="USLEEP"
    (yes that starts with a double underscore).
  • Log out, stop your display manager, restart it.
I'll leave it as an exercise to the reader to figure out what these settings do. (Feel free to explain it in a comment. :) No guarantees of any kind. If this kills kittens you have been warned. Cheers.

Friday, July 3, 2015

KDEPIM without Akonadi

As you know, Gentoo is all about flexibility. You can run bleeding edge code (portage, our package manager, even provides you with installation from git master KF5 and friends) or you can focus on stability and trusted code. This is why we've been offering our users for the last years KDEPIM 4.4.11.1 (the version where KMail e-mail storage was not integrated with Akonadi yet, also known as KMail1) as a drop-in replacement for the newer versions.
Recently the Nepomuk search framework has been replaced by Baloo, and after some discussion we decided that for the Nepomuk-related packages it's now time to go. Problem is, the old KDEPIM packages still depend on it via their Akonadi version. This is why - for those of our users who prefer to run KDEPIM 4.4 / KMail1 - we've decided to switch to Pali Rohár's kdepim-noakonadi fork (see also his 2013 blog post and the code).The packages are right now in the KDE overlay, but will move to the main tree after a few days of testing and be treated as an update of KDEPIM 4.4.11.1.
The fork is essentially KDEPIM 4.4 including some additional bugfixes from the KDE/4.4 git branch, with KAddressbook patched back to KDEPIM 4.3 state and references to Akonadi removed elsewhere. This is in some ways a functionality regression since the integration of e.g. different calendar types is lost, however in that version it never really worked perfectly anyway.

For now, you will still need the akonadi-server package, since kdepimlibs (outside kdepim and now at version 4.14.9) requires it to build, but you'll never need to start the Akonadi server. As a consequence, Nepomuk support can be disabled everywhere, and the Nepomuk core and client and Akonadi client packages can be removed by the package manager (--depclean, make sure to first globally disable the nepomuk useflag and rebuild accordingly).

You might ask "Why are you still doing this?"... well. I've been told Akonadi and Baloo is working very nicely, and again I've considered upgrading all my installations... but then on my work desktop where I am using newest and greatest KDE4PIM bug 338658 pops up regularly and stops syncing of important folders. I just don't have the time to pointlessly dig deep into the Akonadi database every few days. So KMail1 it is, and I'll rather spend some time occasionally picking and backporting bugfixes.

Sunday, June 21, 2015

Perl 5.22 testers needed!

Gentoo users rejoice, for a few days already we have Perl 5.22.0 packaged in the main tree. Since we don't know yet how much stuff will break because of the update, it is masked for now. Which means, we need daring testers (preferably running ~arch systems, stable is also fine but may need more work on your part to get things running) who unmask the new Perl, upgrade, and file bugs if needed!!!
Here's what you need in /etc/portage/package.unmask (and possibly package.accept_keywords) to get started (download); please always use the full block, since partial unmasking will lead to chaos. We're looking forward to your feedback!
# Perl 5.22.0 mask / unmask block
=dev-lang/perl-5.22.0
=virtual/perl-Archive-Tar-2.40.0
=virtual/perl-Attribute-Handlers-0.970.0
=virtual/perl-B-Debug-1.230.0
=virtual/perl-CPAN-2.110.0
=virtual/perl-CPAN-Meta-2.150.1
=virtual/perl-CPAN-Meta-Requirements-2.132.0
=virtual/perl-Carp-1.360.0
=virtual/perl-Compress-Raw-Bzip2-2.68.0
=virtual/perl-Compress-Raw-Zlib-2.68.0
=virtual/perl-DB_File-1.835.0
=virtual/perl-Data-Dumper-2.158.0
=virtual/perl-Devel-PPPort-3.310.0
=virtual/perl-Digest-MD5-2.540.0
=virtual/perl-Digest-SHA-5.950.0
=virtual/perl-Exporter-5.720.0
=virtual/perl-ExtUtils-CBuilder-0.280.221
=virtual/perl-ExtUtils-Command-1.200.0
=virtual/perl-ExtUtils-Install-2.40.0
=virtual/perl-ExtUtils-MakeMaker-7.40.100_rc
=virtual/perl-ExtUtils-ParseXS-3.280.0
=virtual/perl-File-Spec-3.560.0
=virtual/perl-Filter-Simple-0.920.0
=virtual/perl-Getopt-Long-2.450.0
=virtual/perl-HTTP-Tiny-0.54.0
=virtual/perl-IO-1.350.0
=virtual/perl-IO-Compress-2.68.0
=virtual/perl-IO-Socket-IP-0.370.0
=virtual/perl-JSON-PP-2.273.0
=virtual/perl-Locale-Maketext-1.260.0
=virtual/perl-MIME-Base64-3.150.0
=virtual/perl-Math-BigInt-1.999.700
=virtual/perl-Math-BigRat-0.260.800
=virtual/perl-Module-Load-Conditional-0.640.0
=virtual/perl-Module-Metadata-1.0.26
=virtual/perl-Perl-OSType-1.8.0
=virtual/perl-Pod-Escapes-1.70.0
=virtual/perl-Pod-Parser-1.630.0
=virtual/perl-Pod-Simple-3.290.0
=virtual/perl-Safe-2.390.0
=virtual/perl-Scalar-List-Utils-1.410.0
=virtual/perl-Socket-2.18.0
=virtual/perl-Storable-2.530.0
=virtual/perl-Term-ANSIColor-4.30.0
=virtual/perl-Term-ReadLine-1.150.0
=virtual/perl-Test-Harness-3.350.0
=virtual/perl-Test-Simple-1.1.14
=virtual/perl-Text-Balanced-2.30.0
=virtual/perl-Text-ParseWords-3.300.0
=virtual/perl-Time-Piece-1.290.0
=virtual/perl-Unicode-Collate-1.120.0
=virtual/perl-XSLoader-0.200.0
=virtual/perl-autodie-2.260.0
=virtual/perl-bignum-0.390.0
=virtual/perl-if-0.60.400
=virtual/perl-libnet-3.50.0
=virtual/perl-parent-0.232.0
=virtual/perl-threads-2.10.0
=virtual/perl-threads-shared-1.480.0
=dev-perl/Test-Tester-0.114.0
=dev-perl/Test-use-ok-0.160.0

# end of the Perl 5.22.0 mask / unmask block
After the update, first run
emerge --depclean --ask
and afterwards
perl-cleaner --all
perl-cleaner should not need to do anything, ideally. If you have depcleaned first and it still wants to rebuild something, that's a bug. Please file a bug report for the package that is getting rebuilt (but check our wiki page on known Perl 5.22 issues first to avoid duplicates).


Sunday, May 10, 2015

Reviving Gentoo VMware support

Sadly over the last months the support for VMware Workstation and friends in Gentoo dropped a lot. Why? Well, I was the only developer left who cared, and it's definitely not at the top of my Gentoo priorities list. To be honest that has not really changed. However... let's try to harness the power of the community now.

I've pushed a mirror of the Gentoo vmware overlay to Github, see

If you have improvements, version bumps, ... - feel free to generate pull requests. Everything related to VMware products is acceptable. I hope some more people will over time sign up and help merging. Just be careful when using the overlay, it likely won't get the same level of review as ebuilds in the main tree.

Monday, May 4, 2015

PRB accepted: Transport across a carbon nanotube quantum dot contacted with ferromagnetic leads: experiment and non-perturbative modeling

It's time to announce yet another nice result. Our manuscript "Transport across a carbon nanotube quantum dot contacted with ferromagnetic leads: experiment and non-perturbative modeling" has been accepted as a regular article by Physical Review B.
When ferromagnetic materials are used as contacts for a carbon nanotube at low temperature, the current is strongly influenced by the direction of the contact magnetization via the so-called tunneling magnetoresistance (TMR). Since the nanotube contains a quantum dot, in addition its electronic energy levels play an important role; the TMR depends on the gate voltage value and can reach large negative and positive values. Here, in another fruitful joint experimental and theoretical effort, we present both measurements of the gate-dependent TMR across a "shell" of four Coulomb oscillations, and model them in the so-called "dressed second order" framework. The calculations nicely reproduce the characteristic oscillatory patterns of the TMR gate dependence.

"Transport across a carbon nanotube quantum dot contacted with ferromagnetic leads: experiment and non-perturbative modeling"
A. Dirnaichner, M. Grifoni, A. Prüfling, D. Steininger, A. K. Hüttel, and Ch. Strunk
Phys. Rev. B 91, 195402 (2015); arXiv:1502.02005 (PDF)

Wednesday, March 11, 2015

PRB accepted: Broken SU(4) symmetry in a Kondo-correlated carbon nanotube

We're happy to be able to announce that our manuscript "Broken SU(4) symmetry in a Kondo-correlated carbon nanotube" has been accepted for publication in Physical Review B.
This manuscript is the result of a joint experimental and theoretical effort. We demonstrate that there is a fundamental difference between cotunneling and the Kondo effect - a distinction that has been debated repeatedly in the past. In carbon nanotubes, the two graphene-derived Dirac points can lead to a two-fold valley degeneracy in addition to spin degeneracy; each orbital "shell" of a confined electronic system can be filled with four electrons. In most nanotubes, these degeneracies are broken by the spin-orbit interaction (due to the wall curvature) and by valley mixing (due to, as recently demonstrated, scattering at the nanotube boundaries). Using an externally applied magnetic field, the quantum states involved in equilibrium (i.e., elastic, zero-bias) and nonequilibrium (i.e., inelastic, finite bias) transitions can be identified. We show theoretically and experimentally that in the case of Kondo correlations, not all quantum state pairs contribute to Kondo-enhanced transport; some of these are forbidden by symmetries stemming from the carbon nanotube single particle Hamiltonian. This is distinctly different from the case of inelastic cotunneling (at higher temperatures and/or weaker quantum dot-lead coupling), where all transitions have been observed in the past.

"Broken SU(4) symmetry in a Kondo-correlated carbon nanotube"
D. R. Schmid, S. Smirnov, M. Marganska, A. Dirnaichner, P. L. Stiller, M. Grifoni, A. K. Hüttel, and Ch. Strunk
Phys. Rev. B 91, 155435 (2015) (PDF)

Saturday, January 31, 2015

Choice included

Some time ago, Matteo Pescarin created the great "Gentoo Abducted" design. Here are, after some minor doodling for the fun of it, several A0 posters based on that design, pointing out the excellent features of Gentoo. Released under CC BY-SA 2.5 as the original. Enjoy!



PDF SVG


PDF SVG


PDF SVG


PDF SVG


PDF SVG