libera/#sbcl - IRC Chatlog
Search
20:05:31
Krystof
I expect when writing tests for an area that's been basically uncovered by tests for ~20 years to find bugs
20:06:15
Krystof
but unread-char with :io streams (and, admittedly, newline conversion) triggering heap exhaustion...
20:47:58
qhong
I don't have :IMMOBILE-SPACE on *FEATURES*, but (sb-kernel:immobile-space-obj-p #'rplacd) => T, not sure will it move
20:59:55
|3b|
you don't really want /all/ symbols immobile in general, since you will accumulate gensyms indefinitely
21:00:54
qhong
|3b|: Is there a dynamic variable to control it (say during a particular READ) then? Or all INTERNed symbol. I only use INTERNed symbols.
21:14:14
qhong
The only reference my own heap has into SBCL heap is functions and symbols. So just need to make them not move, then it works TM
21:40:54
qhong
Hmm, if now SBCL does not move anything where am I getting this #<unknown immediate object, lowtag=#b1101, widetag=#x6D {E40000006D}> from...
2:20:51
|3b|
The value #<SB-FORMAT::FMT-CONTROL "The result type from ~A:~% ~/SB-IMPL:PRINT-TYPE/~%conflicts with the definition's result type:~% ~/SB-IMPL:PRINT-TYPE/"> is not of type STRING when binding SB-C::FORMAT-STRING
2:21:54
|3b|
(very simple functions that just return a constant, which i assume is what it was trying to say didn't match)
2:32:47
|3b|
ACTION wonders if note-lossage works properly at all, since it seems to be declared to take a string but gets optimized to be passed a compiled format string
2:49:28
|3b|
presumably compiler-mumble and note-unwinnage are similarly broken, since they share the optimization and defknown
2:49:51
|3b|
test case: (declaim (inline foo)) (defun foo () 1) (setf (fdefinition 'foo) (lambda () 2)) (defun bar () (foo))