From: james a. <jam...@se...> - 2004-02-29 20:52:58
|
hello; i'm afraid that i'm at a temporary disadvantage. apple has my osx machine and will likely be keeping it for some time still. i told them i don't want it back until they can figure out why it panics every couple of weeks. we'll see how that turns out. since i'm going to be away next week anyway, that didn't bother to me when i gave it to them. that does, however, make it hard to check your patches. so i'm limited to what i can say from your comments and the diff. Bruno Haible wrote: > > Hi, > > Before the 2.33 release, I'd like to commit a few modifications for the > method combinations handling: > > - Improved error message when a method with invalid qualifiers is added > to a generic function. Don't wait until the computation of the effective > method for this error message. > - Same error message when the method combination of a generic function > changes in a way that makes the old methods invalid. i feel uneasy about these. the phrasing of invalid-method-error and define-method-combination could well leave one with the impression, that the constraint applies to applicable methods only, and thus is applied once a set of applicable methods has been determined. which cannot be until the effective method is to be computed. in addition, as the ultimate decision must be left to the combination, both since arbitrary predicates are supported, and because of *, the benefits are not universal. finally, the error will be seen in connection with a function which does not appear in the standard. what about a warning instead? > - Improved error message when CALL-NEXT-METHOD is called in a :BEFORE or > :AFTER method. We can't do it at DEFMETHOD time, only at runtime, but > anyway it's better than nothing. > - Fix a bug introduced on 2004-02-09: the first argument of > %OPTIMIZE-FUNCTION-LAMBDA was wrong. > - Don't use (COMPILE NIL ...) on code that comes partially from the user. > Otherwise the user cannot debug his own code. was this changed anywhere except w.r.t. the generated lambda which partitioned the methods? > - Comment style. > > The "diff -w" is appended. Comments? other than those above, i'd like to wait until i can take a better look. > > Bruno > > 2004-02-28 Bruno Haible <br...@cl...> > > ... ... |