DEFTYPE does not care about the second argument kind

[Roman Marynchak]

DEFTYPE in SBCL accepts a wide range of garbage to be its second argument, instead of checking it to be a valid lambda-list, as defined by CLHS.

For example:

* (deftype rr yy)
; in: LAMBDA NIL
; (LET* ((YY (CDR #:WHOLE676)))
; (BLOCK RR))
;
; caught STYLE-WARNING:
; The variable YY is defined but never used.
;
; compilation unit finished
; caught 1 STYLE-WARNING condition

Regards,
Roman

[Roman Marynchak (roman-marynchak)]

I have attached the simple patch. Now the behavior is the next:

* (deftype rr yy)

debugger invoked on a SIMPLE-ERROR in thread #<THREAD "initial thread" RUNNING
                                               {A9EB801}>:
  Arguments list is not a valid lambda-list: YY

Type HELP for debugger help, or (SB-EXT:QUIT) to exit from SBCL.

restarts (invokable by number or by possibly-abbreviated name):
  0: [ABORT] Exit debugger, returning to top level.

((LAMBDA (#:WHOLE6 #:ENVIRONMENT7)) (DEFTYPE RR YY) #<unavailable argument>)
0]

Also I have fixed another error message to be a valid English sentence.

Regards,
Roman

[Nathan Froyd (froydnj)]

A testcase in tests/ would be a nice addition. Also, I think the usual phrase is "argument list."

[Roman Marynchak (roman-marynchak)]

Thanks for the note, fixed.

The test is coming soon (I hope to create it till the end of the week).

Regards,
Roman

[Roman Marynchak (roman-marynchak)]

I have attached a simple test. It passes after the fix, and claims about "Unhandled error" without the fix.

Regards,
Roman

[Nikodemus Siivola (nikodemus)]

Thank you. Something like this was merged in 1.0.41.50.

I opted for putting the check in PARSE-DEFMACRO so that other users of that gain the benefit of this as well, and added an utility function for signalling simple-type-errors: there are plenty of places that could use that instead of doing keyword calls...

[Nikodemus Siivola (nikodemus)]

Apropos: we are not currently using the "nominate for trunk" stuff on launchpad.