Error printing for-loop scoping deprecation message

[mbauman]

Didn't this use to print a nice warning or error about the need for a global i inside the for loop body?

julia> i = 1; for _ in 1:1; i += i; end
Exception handling log message: MethodError(print, (IOBuffer(data=UInt8[...], readable=true, writable=true, seekable=true, append=false, size=2, maxsize=Inf, ptr=3, mark=-1), nothing), 0xffffffffffffffff)

[mbauman]

Likely culprit: #27829. cc: @JeffBezanson

[vtjnash]

Why is that throwing a MethodError? Are we miscompiling something? print should have at least some definition for all types.

[yuyichao]

print should have at least some definition for all types.

Not anymore after #27829 .....

[JeffBezanson]

It now throws a MethodError on nothing. We could change that to a different kind of error if we want though.

[JeffBezanson]

What I really want to know is why some logging code is trying to print nothing.

[fredrikekre]

What I really want to know is why some logging code is trying to print nothing.

The module argument sometimes becomes nothing, I guess when it can't be determined for some reason. Incidentally I opened #27792 before print(nothing) was removed. Perhaps that will help a bit here at least? Perhaps it might also be that the file argument becomes nothing?

[JeffBezanson]

Fixed by #27792