Thanks! This is merged now.

"Controversial" is perhaps saying a bit much, my comments were meant more as suggestions for improvement. I like the current pull request, from my side it's good to merge.

Looks like the change was more controversial than I thought. I updated the pull request, backing out of the change to the print statement and the exception, and extracting a function for rounding time limits with some slack.

> They are a bit tangential to the bug but non-controversial I hope.

In my view cleaning up/fixing things related to the code we fix is good, if it doesn't make the issue too complex or hard to review.

I handled the comments. They are a bit tangential to the bug but non-controversial I hope.

I added two comments, but the general fix looks good to me, thanks!

We discussed this on discord and decided to convert all time limits to int as soon as possible:
https://github.com/aibasel/downward/pull/120

If we use --overall-time-limit, our time limits become fractional because a limit for one component is computed as the overall limit minus the elapsed time, which is measured as a float. However, the function setting the limit expects an integer. Up to Python 3.9 floats were automatically converted to ints even though the documentation of the function specified the parameter as an integer. Starting from 3.10, passing a float is considered an error (see https://bugs.python.org/issue43101). For the fix, we simply have to convert our limits to ints. The main question is where to do this.