I’ve been doing that for years. Rollbacks are very rare, to the point that it doesn’t make much of a difference whether I do them all at once or not, other than spending more time to do it.

If I wasn’t using containers for everything, sure. Otherwise it’s a bit of an excessive concern.

exactly my point, I’d suggest automating that before I bothered with PRs that upgrade versions, as it’s a waste of time.

“manual changes”, which connotes “local changes”

It doesn’t. Manual as in a PR with upgrades that you’re suggesting yourself, as opposed to running dependabot.

Putting up a PR with changes isn’t considered a manual anything.

If I have to open a PR myself, that’s very much a manual change.

that’s a lot of FUD, topgrade just upgrades using all package managers you have, it doesn’t do the upgrades itself bypassing the manager that installed it, or package authors.

dependabot is a tool for repos, not to apply local changes

That may work for a handful of projects. It’d be my full time job if I did it for everything I run. Also, I might simply suggest maintainers to adopt dependabot or an alternative before I spend time with manual changes. These things should be automated.

what’s the alternative? Write a PR yourself?

upgrade all things by default

https://github.com/topgrade-rs/topgrade/

squeal

f-s-tab is feeble. Unsatisfactory. Bureaucratic.

f-stab is jocose. Nonchalant. Sharp.

you do have an nginx process with PID 2511847 using the port

get more info with

ps aux | grep 2511847

or kill it, if you need to spawn a new one with the right configuration

find out which process is really binding to 443 if you don’t recognize that port as being used

sudo ss -unapt | grep 443

IMO tensorflow always had a worse API than PyTorch. There’s even the legendary issue “I fucking hate tensorflow”, now unfortunately censored as spam lol.

Tensorflow died because devs never bothered improving that. While PyTorch always had an increasing number of features and high level capabilities, TF has always felt like a lower level tool that only made sense choosing if you needed to run models for inference in other platforms. PyTorch Lightning on top of it was a great touch for researchers.

I wrote an app in 2016 and maintained it for some 3 years. Every year there would be a number of deprecated things that required code changes and it was a pretty simple app. I only imagine the amount of work more complicated apps demand.

This is such an underutilized and neglected behavior.

The very least a config parser should do is to log a warning.

because of garbage like that I always use the long option names in scripts, even when the short one would be obvious

I like the branch names auch

maybe i am a rug

overeducated people who can’t see that “master” has multiple meanings.