Fix #29266, better error message for deprecated scalar-fill indexed assignment

[dourouc05]

No description provided.

[mbauman]

Continuing the discussion from #29266 (comment):

What would be the next steps?

You have test failures, look into the logs (I find the FreeBSD CI service to have the most readable ones) and see what was wrong. Or try running make testall in your own build of Julia. In this case it's an easy fix: just do what it says to fix it.

The next step is to add tests for this specific behavior. Take the failing cases listed in #29266, add the ones you ran into, and put them into a test file — somewhere in test/broadcast.jl would make sense for this change. It doesn't really matter much where you put them within that file — just create a new section there.

[dourouc05]

The last commit adds two test cases for this. If the test suite still passes, I guess everything is OK!

[dourouc05]

I don't get why the tests don't pass on Windows (and the failure seems unrelated), while they do on FreeBSD…

[dourouc05]

That sounds better, indeed!

[dourouc05]

This PR still has the tag "needs tests", but I provided some (fe4e503). Is there anything else needed from my side for this to get merged?

[dourouc05]

@mbauman Is this PR ready to be merged?

[mbauman]

Thanks for the bump! I'm just going to close and re-open to re-run the tests. I don't anticipate any trouble, but it's been some time since the last run. Once they pass, this is good to merge!

[dourouc05]

All bots have all tests running perfectly, except analyzegc and Windows, but these failures are unrelated: on Windows, Julia fails to compile; for analyzegc, it seems a rule is missing in the Makefile.