Postings by Alexander Ploumistos

Concurrent changelogs - merge or nix?


For a few months now and in order to help upstreams find bugs and test
new features I have been maintaining in copr development versions of a
couple of my packages, namely pre-release versions of Molsketch and
development snapshots of SciDAVis built against Qt5. At the same time,
I have been updating the packages in Fedora, when there are new,
stable versions and backporting fixes. As a result, the changelogs
have diverged. When it's time to merge the changes in Fedora, how
should I deal with them?

Qt file dialogs under GNOME and MATE


While ironing out the kinks in the upcoming version of Molsketch with
the upstream developer, we noticed that we weren't seeing the same
dialogs. He is on KDE (OpenSuSE) and I was testing in GNOME and MATE.
Molsketch has an option to save a single molecule/structure with the
.msk extension. Under KDE, the extension appears in the file dialog
and is automatically appended to the file name. This does not happen
under GTK.

He installed Fedora in a VM and ran some tests:

Spam/scam links in bugzilla


A few minutes ago I was notified about a post in a closed bug.

Build failure on ppc64le in rawhide


Yesterday I received a notification from koschei about a build failure
of one of my packages in rawhide:
<a href="" title=""></a>
In the build.log it says that g++ encountered a fatal error, but I
really can't figure out what the problem was. On x86_64 and aarch64
there was no such error. Among the changed dependencies there was qt,
redhat-rpm-config and binutils.

sip-related build failures in rawhide


A few days ago scidavis builds started failing in rawhide. Someone
mentioned here that sip was to blame for a number of such failures and
according to koschei they started to happen after the update from
sip-4.19.9-0.2.dev1805261119.fc29 to sip-4.19.12-2.fc29.

According to the logs[0], sip can't find a file:
sip: Unable to find file "QtCore/QtCoremod.sip"

which results in a number of files not getting generated.

Is this a bug in sip, do I need to change something in my spec file,
or should I contact upstream?

Best regards,


Build fails in rawhide only on x86_64 & i686


I submitted updated builds for scidavis and they failed in rawhide[0]
on x86_64 and i686 with the message:

src/main.cpp:160:5: error: 'int main(int, char**)' causes a section
type conflict with 'const QString operator+(const QString&, const

The update contained mostly fixes for python3.7. About a week ago, the
same code built fine in the f29-python side tag. On my part, the only
difference is that I substituted the patches with the same upstream
commits. Has anything else changed in the meantime?



Taskotron test failures (dist.rpmlint)


I've just submitted an update to cmpfit[0] for f28. The automated tests
show errors and warnings only on x86_64 and armv7hl[1] and the errors have
to do with the omission of ldconfig scriptlets. Aren't we supposed to get
rid of these for f28+?

The warnings are:
cmpfit-devel.armv7hl: W: no-documentation
cmpfit-devel.x86_64: W: only-non-binary-in-usr-lib
cmpfit-devel.x86_64: W: no-documentation
Is there any merit to them?

Upcoming gdouros-*-fonts license change and subsequent complications

Hello all,

I doubt most of you know or care about George Douros' family of fonts,
as almost all of them are quite niche.

Build fails on s390x - complains about missing /usr/lib64/lib*.so* file


I submitted a package for review[0] about an hour ago and when I ran a
scratch build in koji[1] there was a failure with this message in

error: File not found:

I had built previous versions of the same package successfully in the
past, with the last successful build either in late December or early

package blocked for tag f27-updates-candidate


I've been trying to build a new version of a package (scidavis) and it
fails on f27 every time with the message "FAILED: BuildError: package
scidavis is blocked for tag f27-updates-candidate".
<a href="" title=""></a>

Is this because of the Final freeze, or is there some other problem?

Updates for Firefox 57 beta


Now that FESCo has ruled that "firefox 57beta is removed from f25/f26
updates-testing but stays in f27/rawhide", could we at least keep
getting new builds in koji for f25/f26? Judging by the feedback in
bodhi, the various threads here, rhbz and FESCo tickets, there is a
number of users who don't mind sticking with Firefox 57. Going back to
v56 would entail undoing a number of changes and dealing with possible
breakages, only to repeat the whole process in about a month.

Is there a seperate SOP for unblocking a package in koji?


About a couple of weeks ago, I went through the new
adoption/unretirement procedure for scidavis and I also used
fedrepo-req-branch to request that f25, f26 and f27 branches be
created as well.

Combining repo/branch requests with POC changes for orphaned/retired packages

Hello all,

Until last week, I hadn't gone through the new process for requesting
repos and branches or unretiring and unorphaning packages and though I
had read the "WhatHappenedToPkgdb" page, none of the intricacies had
After a previously orphaned and retired package got approved, I filed
a request for the repo, separate requests for each branch and after
another day of not being able to push into the remote, I realized that
I had to file a POC change request as well (my mistake, obviously).
A few days ago there was a thread here, where the reasoning behind not
automating repo creatio

Build failures on armv7hl with "No such file or directory"


I have a package under review[0], that I realized had python scripting
support disabled, so I enabled it.

Is /usr/share/mimelnk used for anything any more?

While cleaning up an old spec file, I came across a desktop file,
containing MimeType information, which is to be placed under

Review swaps


I have a couple of packages waiting for review:

scidavis - Application for Scientific Data Analysis and Visualization
<a href="" title=""></a>

which depends on

liborigin3 - A library for reading OriginLab OPJ project files
<a href="" title=""></a>

Both of them have been in a copr for a while:
<a href="" title=""></a>

I'd appreciate it if any of the people I've done reviews for could
return the favor, but I wouldn't mind reviewing a couple more


Better to bundle a library or package different version than upstream?

Dear all,

About ten days ago I asked a question on this list, but I guess on one
hand it was too specific, while on the other it coincided with people
travelling to Flock or being on vacation. As I really want to get the
package in question back in Fedora, I will ask again and I will try to
be more abstract and concise. For anyone interested, there is a link
to the original post at the end of this message.

Package foo, which depends on library bar for importing a certain type
of files, was retired a couple of releases ago.

SciDAVis and liborigin3 vs liborigin & liborigin2 - bundle or obsolete?


Since its retirement from Fedora, SciDAVis[0] has undergone
significant development and I think it is ready to be re-included in
our package collection. After a few months of private builds that I
distributed among co-workers and friends, I set up a copr[1] and I've
been keeping up with the upstream project.

SciDAVis comes with a bundled copy of liborigin[2], which the upstream
developers helped me unbundle. Its version has been bumped to 3.0.0
internally, although there hasn't been a 3.0.0 release yet.

f27 builds in COPR fail without logs


For the past hour or so, I've been trying to rebuild my copr packages
for f27 and rawhide. While there were some hiccups with rawhide, e.g.
not finding mirrors to download packages, after resubmitting them a
couple of times, all builds were successful.

Default location for addons AppStream metadata has changed

Hello all,

Sometime ago, the default location of AppStream metadata for addons
(*.metainfo.xml files) changed from %{_datadir}/appdata/ to

As far as I can tell that change went mostly unnoticed, given that
most of our packages still install addon metadata in

Unicode 10 support across releases

Hello all,

I am going to update the gdouros-symbola font to the newest version,
which contains Unicode 10 glyphs. I have kind of lost track of things,
so I would like to ask if there is a fedora release that shouldn't or
wouldn't make sense to get the update?


hotness tried to map X to an upstream project, but failed: "Could not determine backend for http://url/project/X"


I've just received about a dozen of messages like the one mentioned in
the subject, accompanied with:

pkgdb_updater updated: upstream_url of X
<a href="" title=""></a>

for a number of gdouros-X-fonts that I updated recently and whose
upstream URL had changed.

I have not added these projects to anitya on purpose, because upstream
regularly restructures their website, the source packages do not
contain any version numbers, sometimes packages get updated silently
and the development takes place offline.
My solution for monitoring what upstream does was a

Regression: FileManager1 D-Bus service mix-up in F25 when multiple DEs are installed

desktop environments that come with a FileManager1 (and probably any
other identical) D-Bus service provider, at least when one of them is
GNOME. This might be considered an oversight on upstream's part, but
it causes troubles downstream and we should somehow deal with it.

A few days ago, I started a thread[1] here, about a problem on a
Fedora 25 machine with multiple DEs, where dolphin would take over
whenever I tried to open a desktop folder in GNOME.

File managers compete for desktop domination in F25

Hello all,

After upgrading a computer with multiple DEs to F25 from F23, folders
on GNOME's desktop are managed by dolphin, which runs daemonized and
cannot be killed, unless I also kill kded5 and kdeinit5. There was
already a similar bug about nemo replacing nautilus on the desktop
after a F24→F25 upgrade:

<a href="" title=""></a>

In both cases, our settings seem to be in order. Does anyone have any
clue as to which component (e.g.

liborigin FTBFS


Yesterday I was notified that liborigin failed to build in rawhide:
<a href="" title=""></a>

For the past couple of days I've been away on a work trip and I won't
be back until the end of next week. Unless this failure messes someone
else's packages, would it be OK to ignore the error until then?

Best regards

bodhi: no email notifications for packages in pending state


For the last couple of days I have been submitting updates (with fedpkg)
for some of my packages in all the branches. Currently they all appear as
pending in bodhi and I have not received any messages about them, even for
the ones that got positive karma.

Context menu icons in Gnome 3.16

I'm messing with an old gtk+ program, that I think it's dead upstream
and I'm trying to see if I can resurrect it. Among other issues, it
should add some entries to nautilus' context menus (which it does) but
I can't get their icons to show. Then I noticed that I'm missing all
of my context menu icons, even though they are enabled in gsettings.
Is this a bug or a feature?

Naming packages when upstream uses dashes in the release version


I have been working on a driver package from a hardware vendor and
besides some patches in the code, I am trying to bring the supplied
(messy) spec file close to our guidelines. (While most of the stuff in
the package relies on open source libraries and projects, there are
some files covered by a commercial license, so it will never be
included in fedora.)

What irks me the most, is that the version of the source package is
3.80-1 and this is reflected in the spec file macros (Version: 3.80,
Release 1), which are used extensively.

AppStream license code for public domain works

Hello everyone,

While writing metainfo files for a few fonts, I noticed that their
creator states on his website:

"Fonts and documents in this site are not pieces of property or
merchandise items; they carry no trademark, copyright, license or
other market tags; they are free for any use."

As far as I can tell, that makes them public domain. Is it the same as
CC0-1.0 or is there another license I should choose?

Captive portal detection on wired connections - bug or feature?

During the last hour, I got three pop-ups on my workstation (which is
connected over Ethernet) with blank windows and a title that read "Captive