Re: Package problem: miscompiled macros

Norbert Koch <viteno(a)xemacs.org&gt; wrote: Except that I'm an idiot. The macro.list file is listed in .cvsignore, for the simple fact that it's so quick and easy to generate using gen-macro-list.awk. So the fact that mine is nearly a year old means that *I* haven't run it in all that time, and says nothing about anybody else. That deserves a smack in the forehead. <*smack!*> About a year ago, we had a rash of reports about miscompiled macro calls. I thought that there must be an automatic way of finding those before package release, and wrote the scripts. That they are AWK scripts you can blame on the fact that I wanted to learn AWK at the time.:-) This is the idea. The gen-macro-list.awk script searches through the package code for every defmacro, defmacro*, defsubst, etc. it can find, and outputs the macro/subst name and file of definition for each. This output is stored in macro.list by convention. When the packages are built, save the build messages somewhere, maybe packages-make-all.err to be parallel with the source distribution. Since the byte-compiler complains about unknown functions, what we want to do is match up those complaints with the list of macro/subst definitions. The match-up is done like this: awk -f find-macro-err.awk < packages-make-all.err That spits out a list of the mataches, and where they are defined. It is not perfect; there is at least one false positive on the list. However, it is a durn sight better than nothing. Hmmmm... The existing stuff in fsf-compat is in files named after the Emacs version. However, combine-after-change-calls is in subr.el in Emacs, and we already have a subr.el. How should I name it? Does it really belong in fsf-compat after all? Me, too, except that I've spent yesterday and today basking in the glow of too much consumed turkey. -- Jerry James, thankful for American holidays http://www.ittc.ku.edu/~james/