assert: reverse diff direction

[ryzokuken]

The current direction for the assert diffs is rather peculiar.
While it’s not the “wrong” way to do it per-se, it’s definitely
inconsistent with the ecosystem and is rather unintuitive.

This PR aims to fix that by reversing the direction.

Fixes: #20897

Checklist

• make -j4 test (UNIX), or vcbuild test (Windows) passes
• tests and/or benchmarks are included
• documentation is changed or added
• commit message follows commit guidelines

/cc @BridgeAR@addaleax@targos

[targos]

I'm surprised no tests have to be updated?

[ryzokuken]

I'd be too, I didn't try running and updating tests yet, but will do that today. That said, I hope the format looks good to you?

[targos]

SGTM

[BridgeAR]

-1. This change is not something we should do without having a proper longer discussion about it. The original implementation had a reason and there are quite a few factors to anticipate.

[addaleax]

@BridgeAR I opened an issue for that very reason. Could you weigh in there? So far me + 4 other people feel this way and nobody has disagreed yet (it’s only been open for 2 days, though).

If you say “the original implementation had a reason and there are quite a few factors to anticipate” it would be great to know that reason and those factors :)

[BridgeAR]

I am still tying ;-)

[BridgeAR]

In general I am not -1 on the actual change. More like a +-0. I just think that we should all agree on it first in the issue. Besides the fact that this PR is not even remotely done (there are lots of tests that are going to fail because of such a change).

[ryzokuken]

@BridgeAR agreed. Will make the required changes to our testing suite as soon as I can :)

[addaleax]

Hm. Ping @ryzokuken?

A friend just ran into this issue (assert diffs being the “wrong” way around) and pinged me about it 😕

[BridgeAR]

@addaleax that change is included in #21628. There are more changes in there though.

@ryzokuken I can remove that part again if you would like to land this PR first. But in that case I would like to also change the color and the tests have to be fixed. I could then rebase accordingly.

[ryzokuken]

@BridgeAR agree with the additional changes you have listed. Let's come up with a solid list of TODOs to take care of before landing this and I'd gladly make the changes promptly.

[ryzokuken]

Closing in favor of #21628, which already addresses the said issue.