freenode/#sbcl - IRC Chatlog

14:00:48 jdz I'm still looking into this, but in the meanwhile maybe stassats knows what's up: https://gist.github.com/jdz/13c002e186f2e7f45f3912eb007c65b2

14:03:03 jdz CL-WHO does a bit of rebinding, like (let ((stream stream)) ...), but looking at WITH-HTML-OUTPUT macro (and the MACROLETs) I don't immediately see how to reduce the test case.

14:41:01 jdz Interestingly, removing (check-type ,var stream) from WITH-HTML-OUTPUT macro seems to fix this.

14:51:34 jdz Updated the gist with the shortest test cases.

14:54:11 stassats jdz: it's a dead-code elimination problem

14:56:49 jdz Interestingly (for me) the problem also seems to only affect the FORMAT call, not PRINC.

14:59:04 jdz Although I'd guess (format stream "~A" "string") would be transformed into (princ "string" stream).

14:59:22 stassats it's not related to format or printing or anything

15:00:02 jdz I was thinking compiler macros, but I really don't think I'm adding any information here.

16:33:09 stassats just needs to look through optional-dispatch, but the entry point is in fact called, from the main fun

16:33:17 stassats how com find-dfo removes it anyway

16:37:01 stassats i suppose i see, ok then, no more questions

17:00:42 stassats see even more clearly now, component-head links to any function

17:01:11 stassats but deleting an unused optional-dispatch breaks the link

17:01:35 stassats there still may be a better solution

17:05:27 stassats very useful bug

18:37:18 stassats my own recent test case no longer works...

18:37:22 stassats lp 1862475

18:37:23 specbot https://bugs.launchpad.net/bugs/1862475

18:37:53 stassats neither does this

18:38:25 stassats have to revert two changes to trigger it

19:09:45 stassats jdz: fixed

19:27:05 stassats why didn't pfdietz catch it

19:28:04 stassats it: (lambda (x) (flet ((z (&optional a) 123)) (let ((z #'z)) (when x (unless x (setf z 10))) (funcall z))))

19:28:36 stassats (when x (unless x)) is useful to delay the removal until constraint propagation

19:42:54 pfdietz The (main) random tester limits itself to (mostly) forms that have type INTEGER.

19:44:20 pfdietz One big reason for this is to make forms easy to prune.

19:45:15 pfdietz I'm not surprised there are blind spots.