Python 3.7: Smoother than expected
As I wrote in another post a few weeks ago, Gentoo has officially made the switch. The new default Python version is Python 3.7! I was not totally sure just what to expect with this update. Surely some things would go wrong, and I would encounter some issues somewhere. As it turns out, I was wrong.
All week, even the week before, I have been anticipating with excitement the new default target to switch. I could have made the switch early without too much difficulty, I believe. I wanted to wait it out and get the full experience here. Part of that reasoning is because I knew I had some packages which have not been updated to Python 3.7 yet. Mostly though, I wanted to see how Portage would handle this on it’s own with minimal intervention. And so I waited.
Early in the week, around Monday or Tuesday, readline also received a major
version bump. Due to how many packages rely on readline, the system library
status of the package, I was in for quite a few rebuilds. While not a problem
on it’s own, I noticed quite a few Python packages which would be rebuilt as
part of that readline update. This was not going to sit well with me. Why
spend the time to update several packages, only to rebuild them in another day
or two with a new Python version? And so I waited some more.
Wednesday: Python day!
Wow, was I excited when I woke up Wednesday morning! Had the day off work (for
other reasons — not specifically Python related) so the whole day is mine for
updating and shenanigans. Unfortunately, when I ran my first eix-diff of the
day, nothing. Tried a little portageq action as well, to check the actual
values of PYTHON_TARGETS and PYTHON_SINGLE_TARGET to no avail. The
thought did not occur to me until then that there was no set schedule for the
profile update. It will be available when it’s ready.
Not going to let that get me down though. I found other ways to occupy my time.
Namely a whole lot of Witcher 3. Thankfully, that game was more than enough
to pass the time. And could even be said to be the main reason I ended up
running the actual update the following day. Either way, I sneaked in a couple
extra emerge sync throughout the day, waiting to see when the new targets
were actually pushed out.
Later in the afternoon, they finally came, and initially I was presented with
a whole slew of errors. Unfortunately, I did not think to save them. Mostly
they were errors about inability to build certain packages with the new targets.
Some of my installed packages had not yet been updated to 3.7. Due to the fact
that those particular packages were not things I use regularly, or really at
all anymore, the fix was simple. Remove them. Just a couple packages to remove
and depclean and we’re in business.
Just a little more back and forth with that process until Portage finally stopped complaining, and I was presented with a quite large list of packages to rebuild or update. By this time, it was already nearing 10pm, and with work starting in another 8 hours, it was time for bed. Though I was glad knowing I had already handled the bulk of the hard work and can simply let it run after work.
Thursday: Make it so
Little fast forward action and we’re back at home. Double, and triple check
once more. Make sure everything looks sane. Only 126 packages on my list. No
big deal. emerge @world — get it going!
My original estimate on the total build time was about 3 hours, possibly up to
4. Much to my surprise, the @world update took exactly two hours. Good
stuff! During a couple packages, I was quite a bit anxious. A friend of mine
mentioned having a lot of package slot errors, and some trouble getting
Portage to update at one point. I’m not sure what the cause of their errors
was, but I (very thankfully) did not encounter any.
With the update complete, I went ahead and updated the default Python version
through eselect python, and verified the new version loads properly. Hit
up a emerge @preserved-rebuild as well, which required the Python 3.6 to
be rebuilt against the new readline. Not sure why this did not happen in
place with the others, but it went just fine. Finally, the small process to
restart several daemons/services running which depend on either readline or
Python, followed by a re-login to my X session and we’re almost through. My
last set of things to check was my own Python projects. With the exception of
one, which was not properly using a virtual environment, everything handled
the update flawlessly. Color me impressed.
Thank you, Gentoo Python Team
In the end, this was a fairly large update, with a lot of behind the scenes work by the Gentoo Developers, the Gentoo Python Team, and of course the package maintainers and developers of the actual packages which use Python. I have noticed a huge uptick in Python related package work over the past couple of months, at least. This work has not gone unnoticed, and I am very thankful to have such a great team working to continue improving Gentoo.
Very big thanks goes out to the maintainers and developers who helped to ensure this process went smooth and easy as best they could. No major version update is going to be without problems, especially if the package being updated is part of the “core” required packages by that distribution. The work put in over the past couple months has done a great deal in helping to ensure there were as few issues as possible. Thank you all for the hard work and dedicated. I truly do appreciate the effort and care put into your work. Thank you for helping to make Gentoo the one Linux distribution that truly feels like “home” to me.