Thanks @savannahostrowski for the PR 🌮🎉.. I'm working now to backport this PR to: 3.13, 3.14.
🐍🍒⛏🤖

GH-143588 is a backport of this pull request to the 3.14 branch.

GH-143589 is a backport of this pull request to the 3.13 branch.

There is a call to reporthook just a few lines above before reading the first block. Would it not weird?

Also the docs say

The callable will be passed three arguments; a count of blocks transferred so far, a block size in bytes, and the total size of the file.

I am not sure that passing the kength of the block that was just read is the 'block size' in this case. Or maybe we should documenty it better.

We should also document for blocknum == 0 that we first give the blocksize.

After thinking about this more, I'm inclined to revert this PR. While the old behavior was technically inconsistent, it's been working fine for people, and as @picnixz points out, there are still more edge cases to consider. The change in mental model probably isn't worth the breakage.

Also cc: @hugovk for his 2 cents.

There is much of context to bring, I will do a re-review too.

No strong opinion either way from me. If we do revert it, let's do so before tomorrow's 3.15a4.

Let's keep the behavior same, that is revert this PR, and close the old bug.

We actually didn't want len(block) and wanted a constant bs

https://hg.python.org/cpython/rev/8c48eb0239ca

1. As @picnixz pointed out, anyone reading docs carefully will expect block size the a constant.
2. Would have come to realize this bs behavior since Python 3.3 and we don't want to break that un-necessarily.

Sorry, @savannahostrowski - reverting seems better to me.