Resolve binding MethodError directly on 1.12

[vchuravy]

MethodError is pulled in implicitly and we need to disambiguate our usage here

x-ref: JuliaLang/julia#57290

ought to fix the GPUCompiler issue in JuliaLang/julia#57332

[maleadt]

I'd rather do the other way around; to avoid invalidations we don't really want to expose this method globally, do we? We could just rename it to MethodErrorByType.

[vchuravy]

to avoid invalidations we don't really want to expose this method globally, do we?

IIUC Keno's comment in JuliaLang/julia#57290 (comment) by resolving the binding directly with global MethodError we choose the behavior we want here, e.g. we create a new binding for MethodError that is not aliased with Base.MethodError.

[codecov]

Codecov Report

All modified and coverable lines are covered by tests ✅

Project coverage is 70.21%. Comparing base (4705aea) to head (ad824e3).
Report is 1 commits behind head on master.

Additional details and impacted files

@@            Coverage Diff             @@
##           master     #664      +/-   ##
==========================================
- Coverage   71.71%   70.21%   -1.51%     
==========================================
  Files          24       24              
  Lines        3330     3330              
==========================================
- Hits         2388     2338      -50     
- Misses        942      992      +50     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[topolarity]

Suggested change

	global MethodError
	function MethodError end

imo, it's more conspicuous that this does what you want

[vchuravy]

Actually that looks more confusing since it seems redundant to the statement below.

In every other situation I would merge a function declaration and a method declaration into one.

[topolarity]

Fair enough - I agree that neither syntax is really very transparent / consistent