Fix bugs in definitions of several Float16 methods (ref #17148).

[pkofod]

This PR fixes the bug in

julia/base/float16.jl

Line 157 in 861100c

$func(a::Float16,b::Float16) = Float16($func(Float32(a),Float32(a)))

where the first positional argument is accidentally used in several Float16 methods. Fixes #17148

[vtjnash]

lgtm

[pkofod]

Maybe this should even be tested as well :)
edit should I just add a test for the exact case in #17148 ? @kshyatt did you find this while working on tests for it? In that case you might have tests you want to add yourself, and I should just keep the tests added here to a minimum.

[yuyichao]

.... Please have a better commit message AND PR title AND PR body. None of those mention what this PR is about at all!

[kshyatt]

@pkofod I'm not planning on adding more tests. This LGTM but @yuyichao makes a good point as well.

[yuyichao]

Also note that it is important to have the first line of the commit message be a summary of the change since it's what's used in shortlog and (most importantly) blame.

[pkofod]

.... Please have a better commit message AND PR title AND PR body. None of those mention what this PR is about at all!

Also note that it is important to have the first line of the commit message be a summary of the change since it's what's used in shortlog and (most importantly) blame.

My bad. Used the github interface (hence the patch-2 branch name) since it was such a simple change, and might have clicked create pull request a bit too fast :) I hope it's better now.

[simonbyrne]

LGTM: @kshyatt does this satisfy your testing requirements, or do we need to test them all?

[kshyatt]

This is ok with me. It's not there's some quota on PRs - we can fix it later and it's not like people were getting constantly owned by this.

[simonbyrne]

Thanks!