16:51:28drmeisterI just created an (ext:rmtree "path") function that deletes the file "path" or if "path" is a directory it wipes out everything in the directory and the directory.
16:51:34drmeisterIs this an insanely dangerous thing?
16:52:07drmeisterI needed it for removing old fasl directories.
16:54:27drmeisterBike: I just pushed the fasldir branch.
16:54:34drmeisterIt allows fasls to be directories.
19:17:58Bikedrmeister: here's the diff of my hack "fix" to ld64 http://ix.io/23I9
19:25:54Bikeextensible sequences are starting to work.
19:26:21Bikethe cando stuff might have a little of an issue in that things will be weird if the "sequence" isn't an actual subclass of SEQUENCE. but SEQUENCE is an abstract class so we can probably fake it if nothing else
21:04:10drmeisterMartin suggested we limit the depth that rmtree will work on. So I'll restrict it to a flat directory.
21:04:23drmeisterI can also look for 'fasl' in the name
23:32:20Bikecould put an extra slot in standard-generic-function, its invalidated-function. whenever the generic function would be invalidated, its discriminating function is set to its invalidated-function. for most functions this would be the usual (i.e. a function that recomputes and installs the discriminating function). for whatever system functions, it's a function that handles standardly classed arguments (e.g.
23:32:26Bikefor FIND, (function list)) and only recomputes and installs a new function on a miss.
23:32:32Bikeer. wrong channel. whatever, close enough
23:32:45Bikethe point being that system functions are always usable on standard arguments
23:32:53Bikei guess these could just go in a hash table since this is slow path