Search
Thursday, 17th of October 2019, 11:29:34 UTC
11:38:29
stassats
amelas: hi, do you have any questions?
11:46:51
amelas
I was looking at the tests after building, and I was wondering if there was any documentation for the with-test macro
11:55:09
amelas
I see a lot of tests that are :broken-on :win32
11:55:30
amelas
is there a distinction between :skipped-on and :broken-on?
12:12:20
stassats
amelas: broken-on is reported differently
12:12:38
stassats
skipped-on means it's not applicable
12:13:04
stassats
fails-on means that it can run without hosing the whole thing but doesn't work correctly
12:14:19
amelas
and is there a higher-level meaning? Should tests that are broken-on in theory be fixed?
14:58:41
flip214
I've got a docstring fix at https://github.com/phmarek/sbcl/commit/9ffa314ce37d88e9458d19ec4bd2a5f22b1711ad
15:02:25
stassats
format it and paste/mail it?
15:04:41
flip214
If you want... I thought that (once) registering that remote and then just "git remote update" and "git merge" would be easier
15:06:59
flip214
with appropriate aliases it might be ;)
17:43:03
flip214
I found some more typos... mail on the way.
17:45:46
flip214
stassats: are you planning to attend sbcl20 as well?
18:30:57
pfdietz
I had briefly considered going, but expense and time constraints prevented me.
18:33:59
flip214
pfdietz: ask Krystof about travel assistance
18:39:26
stassats
better to spend the time working on sbcl
18:40:13
stassats
BACKTRACE INTERRUPTED-CONDITION-WAIT keep intermittently failing on sb-safepoint
19:18:07
stassats
of course can't reproduce locally
21:15:48
pfdietz
Hmm. There appears to be untapped potential to optimize various built-ins based on argument types. Sample ticket submitted.
21:18:01
pkhuong
re interruption and synchronisation primitives, I'm beginning to think that the only portable way (i.e., outside of just assuming we have something like futexes, which are available on linux, windows, many BSDs, and, apparently, hidden in OS X) to satisfy our reentrancy/atomicity requirements is to use some form of message passing to a background thread that doesn't get interrupted for GC, etc.
21:28:17
stassats
pfdietz: equal to string=, string= does more stuff than just strings and will error otherwise
21:29:00
stassats
the first thing, it's a simple change from OR to AND in the transform
21:29:07
stassats
will push after testing
21:35:58
stassats
pkhuong: we need better synchronization of any kind on darwin
21:36:05
stassats
a spinlock with a backoff is no good
21:45:29
stassats
pfdietz: could have string-equalp
21:46:14
stassats
(defun f (x) (when (equalp "abc" x) x)) doesn't get derived as STRING
21:46:40
stassats
probably not a big deal, as it's not simple-string, but could aid some bad-code detection
21:49:14
stassats
put it into my notes, will deal later, should be trivial
21:50:07
stassats
i have a large amount of notes on all kinds of possible optimizations, improvements, bad code detectors
Thursday, 17th of October 2019, 23:29:34 UTC