Walkback when compiling a method with an empty category

[ericwinger]

If you try to compile a method with an empty category, you get a walkback. My preference would be that Rowan just puts it in a default category but I'm open to other ideas. The walkback is a pretty harsh failure if the method to compile is otherwise ok.

2025-07-10_53858451_rowanServerStackDump.txt

The "offending" code is here:

RwPrjBrowserToolV2>>addOrUpdateMethod:inProtocol:forClassNamed:isMeta:
<....>
	| methodDef loadedProject loadedPackage couldBeHybrid hybridLoadedPackage hybridLoadedProject loadedClass |
	couldBeHybrid := (hybridPackageName at: 1) = $*.
<....>

[dalehenrich]

GemStone throws an error if you try to pass a nil in as the categoryString using the method Behavior>>compileMethod:category:environmentId:. However, GemStone does provide a single alternate method that doesn't have a category argument: Behavior>>compileMethod:

I'm concerned that "allowing nil category strings" puts the onus on Rowan to protect all of the entry points from nil category strings ... including nils in the tonel data files on disk ...

So I'd be inclined to follow GemStone's model and provide the method Behavior >> rwCompileMethod: as the sole entrypoint for using the default method category '(as yet unclassified)' ... then within Rowan itself (as in GemStone) the assumption can be made that all methods have a category (String) and all methods will have a category string when written to disk ...