Martin S, thanks for the report. I apologize for not getting to this earlier, but I don't know all that much about the byte compiler yet. And I didn't want to waste Martin Buchholz's time, although he's the obvious consultant (and it turns out that he probably made the relevant change); he's on leave from the project at the moment. Martin> Consider an elisp file containing this: (eval-when-compile (defvar x (cl-macroexpand-all '(eval-when-compile (error "evaluated")))) nil) Martin> If this is byte compiled in XEmacs 21.4.6, the error Martin> "evaluated" is signalled inside the byte compiler. It Martin> turns out that the content of the `eval-when-compile' is Martin> evaluated during the `cl-macroexpand-all' call. I did not Martin> expect `error' to be evaluated at that point since it's a Martin> function and not a macro. Martin> This is a regression from (at least) XEmacs 20.4 and Martin> 19.16, where x is simply assigned the quoted value (error Martin> "evaluated") at compile time. FSF Emacs also works that Martin> way. As does XEmacs 21.1. This looks like it's due to a change by Martin Buchholz: 2000-11-17 Martin Buchholz <martin(a)xemacs.org&gt; * bytecomp.el (byte-compile-eval): New. (byte-compile-initial-macro-environment): Use byte-compile-eval. Keeps this promise made in Lispref: "If a file being compiled contains a `defmacro' form, the macro is defined temporarily for the rest of the compilation of that file." I've tried putting the call to byte-compile-top-level back in (see partial diff 21.4 vs 21.1 below); that doesn't help. Changing byte-compile-eval back to eval does help. Martin B, do you have any idea what's going on here? Also, I don't understand what byte-compile-eval is supposed to protect against; I can't find anything---except the initialization and a couple places that bind byte-compile-macro-environment to nil---that actually changes byte-compile-macro-environment. I'm adding Martin S's test case to the regression suite; do you have one that exercises byte-compile-eval? Index: lisp/bytecomp.el =================================================================== RCS file: /pack/xemacscvs/XEmacs/xemacs/lisp/bytecomp.el,v retrieving revision 1.10 retrieving revision 1.11.2.1 diff -u -r1.10 -r1.11.2.1 --- lisp/bytecomp.el 17 Nov 2000 03:21:22 -0000 1.10 +++ lisp/bytecomp.el 25 Jul 2001 07:44:22 -0000 1.11.2.1 ＠＠ -443,16 +443,33 ＠＠ (defvar byte-compiler-error-flag) +;;; A form of eval that includes the currently defined macro definitions. +;;; This helps implement the promise made in the Lispref: +;;; +;;; "If a file being compiled contains a `defmacro' form, the macro is +;;; defined temporarily for the rest of the compilation of that file." +(defun byte-compile-eval (form) + (let ((save-macro-environment nil)) + (unwind-protect + (loop for (sym . def) in byte-compile-macro-environment do + (push + (if (fboundp sym) (cons sym (symbol-function sym)) sym) + save-macro-environment) + (fset sym (cons 'macro def)) + finally return (eval form)) + (dolist (elt save-macro-environment) + (if (symbolp elt) + (fmakunbound elt) + (fset (car elt) (cdr elt))))))) + (defconst byte-compile-initial-macro-environment - (purecopy - '((byte-compiler-options . (lambda (&rest forms) - (apply 'byte-compiler-options-handler forms))) - (eval-when-compile . (lambda (&rest body) - (list 'quote (eval (byte-compile-top-level - (cons 'progn body)))))) - (eval-and-compile . (lambda (&rest body) - (eval (cons 'progn body)) - (cons 'progn body))))) + '((byte-compiler-options . (lambda (&rest forms) + (apply 'byte-compiler-options-handler forms))) + (eval-when-compile . (lambda (&rest body) + (list 'quote (byte-compile-eval (cons 'progn body))))) + (eval-and-compile . (lambda (&rest body) + (byte-compile-eval (cons 'progn body)) + (cons 'progn body)))) "The default macro-environment passed to macroexpand by the compiler. Placing a macro here will cause a macro to have different semantics when expanded by the compiler as when expanded by the interpreter.") -- Institute of Policy and Planning Sciences http://turnbull.sk.tsukuba.ac.jp University of Tsukuba Tennodai 1-1-1 Tsukuba 305-8573 JAPAN My nostalgia for Icon makes me forget about any of the bad things. I don't have much nostalgia for Perl, so its faults I remember. Scott Gilbert c.l.py