State of Calibre in Debian
To counter some recent FUD spread about Calibre in general and Calibre in Debian in particular, here a concise explanation of the current state.
Many might have read my previous post on Calibre as a moratorium, but that was not my intention. Development of Calibre in Debian is continuing, despite the current stall.
Since it seems to be unclear what the current blockers are, there are two orthogonal problems regarding recent Calibre in Debian: One is the update to version 4 and the switch to qtwebengine, one is the purge of Python 2 from Debian.
Debian sid and testing currently hold Calibre 3.48 based on Python 2. Due to the ongoing purge, necessary modules (in particular python-cherrypy3) have been removed from Debian/sid, making the current Calibre package RC buggy (see this bug report). That means, that within reasonable time frame, Calibre will be removed from testing.
Now for the two orthogonal problems we are facing:
Calibre 4 packaging
Update 20191028: Calibre 4.2 based on Python2 has been uploaded to unstable due to availability of the necessary modules.
Calibre 4 is already packaged for Debian (see the master-4.0 branch in the git repository). Uploading was first blocked due to a disappearing python-pyqt5.qwebengine which was extracted from PyQt5 package into its own. Thanks to the maintainers we now have a Python2 version build from the qtwebengine-opensource-src package.
But that still doesn’t cut it for Calibre 4, because it requires Qt 5.12, but Debian still carries 5.11 (released 1.5 years ago).
So the above mentioned branch is ready for upload as soon as Qt 5.12 is included in Debian.
The other big problem is the purge of Python 2 from Debian. Upstream Calibre already supports building Python 3 versions since some months, with ongoing bug fixes. But including this into Debian poses some problems: The first stumbling block was a missing Python3 version of mechanize, which I have adopted after a 7 years hiatus, updated to the newest version and provided Python 3 modules for it.
Packaging for Debian is done in the experimental branch of the git repository, and is again ready to be uploaded to unstable.
But the much bigger concern here is that practically none of the external plugins of Calibre is ready for Python 3. Paired with the fact that probably most users of Calibre are using one or the other external plugin (just to mention Kepub plugin, DeDRM, …), uploading a Python 3 based version of Calibre would break usage for practically all users.
Since I put our (Debian’s) users first, I have thus decided to keep Calibre based on Python 2 as long as Debian allows. Unfortunately the overzealous purge spree has already introduced RC bugs, which means I am now forced to decide whether I upload a version of Calibre that breaks most users, or I don’t upload and see Calibre removed from testing. Not an easy decision.
Thus, my original plan was to keep Calibre based on Python 2 as long as possible, and hope that upstream switches to Python 3 in time before the next Debian release. This would trigger a continuous update of most plugins and would allow users in Debian to have a seamless transition without complete breakage. Unfortunately, this plan seems to be not actually executable.
Now let us return to the FUD spread:
- Calibre is actively developed upstream
- Calibre in Debian is actively maintained
- Calibre is Python 3 ready, but the plugins are not
- Calibre 4 is already in Debian unstable (Python2) and experimental (Python3)
Calibre 4 is ready for Debian as soon as the dependencies are updated
- Calibre/Python3 is ready for upload to Debian, but breaks practically all users
Hope that helps everyone to gain some understanding about the current state of Calibre in Debian.