Re: [AC21.5] Fix a nasty performange regression in syntax cache

robert delius royar writes: You're very welcome. If you find yourself "not getting around to it" twice, just submit whatever you can remember offhand. We know you usually take care, and "making a good bug report" is only Rule #2. Rule #1 is "make the report"!:-) [Aside to first-time reporters: the same advice goes for you, but if you make too many undiagnosable reports without making several "good" reports as well, your reputation will be harmed and your reports will get low priority.] I believe that in modern XEmacsen profiling is always enabled. So you can do "M-x profile-command RET font-lock-fontify-buffer RET"[1] and attach the output. Don't worry about error-checking configurations (if error-checking is enabled, the output warns that it may be inaccurate due to error-checking). For a performance *regression* accuracy in ticks/function is not important.[2] What we'll be looking for in the profiling output is anomolies in the number of function calls, especially expensive ones. These are recorded accurately by Ffuncall itself, and in a few hot spots at the C level as well. Footnotes: [1] Or "M-x profile-key-sequence RET C-x C-f deliver.c RET" for that matter. If you can repeatedly cause XEmacs to perform badly, just use profile-command or profile-key-sequence on the last thing in the recipe. Narrowing down the regression as much as possible will probably be done by the profiler itself! [2] For optimization requests ("this has always been annoyingly slow, please speed it up"), a non-error-checking build is important. There's no point in optimizing something that only takes 0.5% of the ticks normally, but sometimes error-checking can promote that to 5% or even 10%! _______________________________________________ XEmacs-Beta mailing list XEmacs-Beta(a)xemacs.org http://lists.xemacs.org/mailman/listinfo/xemacs-beta