EVAL occasionally expands macros twice

[Richard Holcombe]

I execute the following code:

(defvar *arglist* nil)
(defmacro test2(x)
(progn
(push x *arglist*)
`(setf ,x 7)))

(test2 y)

then *arglist* has the value (y y) when just (y) is expected. (at least (y) is what I get on all of the other lisps that I can find, and makes much more sense.)

This occurs at least on versions 1.0.33 and 1.0.36 on linux.

uname -a
Linux 2.6.31-20-generic #58-Ubuntu SMP Fri Mar 12 05:23:09 UTC 2010 i686 GNU/Linux

*features*

(:ANSI-CL :COMMON-LISP :SBCL :SB-DOC :SB-TEST :SB-LDB :SB-PACKAGE-LOCKS
 :SB-UNICODE :SB-EVAL :SB-SOURCE-LOCATIONS :IEEE-FLOATING-POINT :X86 :UNIX :ELF
 :LINUX :SB-THREAD :LARGEFILE :GENCGC :STACK-GROWS-DOWNWARD-NOT-UPWARD
 :C-STACK-IS-CONTROL-STACK :COMPARE-AND-SWAP-VOPS :UNWIND-TO-FRAME-AND-CALL-VOP
 :RAW-INSTANCE-INIT-VOPS :STACK-ALLOCATABLE-CLOSURES :STACK-ALLOCATABLE-VECTORS
 :STACK-ALLOCATABLE-LISTS :STACK-ALLOCATABLE-FIXED-OBJECTS :ALIEN-CALLBACKS
 :CYCLE-COUNTER :INLINE-CONSTANTS :LINKAGE-TABLE :OS-PROVIDES-DLOPEN
 :OS-PROVIDES-PUTWC :OS-PROVIDES-SUSECONDS-T)

[Nikodemus Siivola (nikodemus)]

Strictly speaking IIRC, macros are allowed to be expanded multiple times, what is going on here seems unintentional.

[Douglas Katzman (dougk)]

This behavior is a consequence of *evaluator-mode* being :compile.
The REPL tries to use an extremely rudimentary evaluation strategy to avoid calling the compiler.
The first thing that it has to do to try interpreting a form is to macroexpand it, accounting for one call of TEST2.

For a variety of reasons the simplistic evaluator will give up, and call the compiler, which it does using the original form.
In this case it's a SETQ of an unknown variable that is not handled by the simple evaluator, so it then calls the compiler on the original form, which has to be expanded again.
If Y had been DEFVAR'ed, the simple evaluator would have done the SETQ and not called the compiler.

As Nikodemus remarked, this is permissible behavior. With *evaluator-mode* = :interpret, only one macroexpansion occurs.
Basically not a bug.