Search
Thursday, 12th of December 2019, 11:24:50 UTC
13:28:06
drmeister
Still no official word from the DoD about our new endeavor. But they are asking me to explain things with spreadsheets - soon...
15:35:13
drmeister
Bike: I just pushed the change to the scraper that catches problems with &rest/core:&va-rest
15:36:02
Bike
like i said, once i just changed it it worked, like you thought
18:44:44
kpoeck
Bike core::closest-sequence-type no longer exists, but is used in the regression-tests (closest-sequence-type0 and closest-sequence-type1)
18:45:17
kpoeck
Is core::sequence-type-maker-info the new equivalent?
18:45:50
Bike
what's using it in the tests?
18:46:44
kpoeck
(test closest-sequence-type0 (eq (core::closest-sequence-type 'simple-base-string) 'base-char))
18:46:54
kpoeck
and (test closest-sequence-type1 (eq (core::closest-sequence-type 'simple-string) 'character))
18:47:12
kpoeck
(test closest-sequence-type0 (equal (core::sequence-type-maker-info 'simple-base-string) '(vector base-char)))
18:47:17
Bike
wonder if we couldn't test coerce or concatenate or something less directly
18:47:23
Bike
but yeah, that's fine i guess
18:47:23
kpoeck
(test closest-sequence-type1 (equal (core::sequence-type-maker-info 'simple-string) '(vector character)))
18:47:59
kpoeck
it is now (vector character) instead of 'character, but looks fine for me
18:49:21
Bike
since we have the sequences extension now it sometimes returns classes
19:43:10
drmeister
I’m heading back to temple after having lunch with someone from GSK
19:44:19
drmeister
Bike. I hated the idea of scratching our heads a year from now if that mistake is made again
19:44:36
drmeister
Defensive programming
20:25:12
kpoeck
I just run the ansi-tests again with very underwhelming results
20:25:53
kpoeck
There are 191 unexpected failures:
20:26:32
kpoeck
72 with sequences, 62 with structures
20:27:14
kpoeck
Two tests result in a segmentation violation
20:28:23
kpoeck
I assume that Martin did not integrate the test in the continous delivery
20:28:34
Bike
well, it's probably just a few stupid things causing a bunch of problems
20:28:48
Bike
....i didn't do anything with structures, though
20:28:59
Bike
oh, i guess it might be going through elt or something. what ever
20:29:17
Bike
could you pick a sequence test failure at random for me to look at, please?
20:29:28
kpoeck
Yes, probably few thing that causes a lot of errors
20:29:44
kpoeck
(Search "1" "212") now returns nil
20:30:39
kpoeck
Most of the other tests seem to have to do with error situations
20:31:54
kpoeck
e.g. (COPY-SEQ 10) now results in NO-APPLICABLE-METHOD-ERROR, used to be a type-error
20:31:59
Bike
oh, wait, that's search-string, i did... something, there
20:32:08
Bike
oh, those. yeah, sure, i just need to throw in some more methods there
20:33:02
kpoeck
I believe there are like 50 tests that used to raise a type-error and now result in NO-APPLICABLE-METHOD-ERROR
20:34:06
Bike
search string is an actual problem, though, as is this segfault business
20:34:51
kpoeck
the segfault I believe i understood, this is because of safety 0 and the cons declaration
20:35:05
kpoeck
and it might not be a cons
20:35:55
kpoeck
(CONCATENATE 'LIST) seems to should return nil, but results in NO-APPLICABLE-METHOD-ERROR
20:36:24
kpoeck
Well let me find representatives of the errors and put them in issues
20:40:54
Bike
oh, the concatenate one is dumb too
20:50:18
kpoeck
The struct error seem all to be with illegal types
20:50:25
kpoeck
e.g. (defstruct (struct-test-51 (:type (vector (integer 0 255)))))
20:50:37
kpoeck
(VECTOR (INTEGER 0 255)) is not a valid :TYPE in structure definition for STRUCT-TEST-51
20:51:00
Bike
oh! right, i did change that, i forgot
20:51:32
Bike
wait, how is that illegal?
20:52:30
Bike
i mean, that's okay in the language, right? it's clasp being dumb
20:57:36
kpoeck
CLHS question. In http://www.lispworks.com/documentation/lw50/CLHS/Body/m_defstr.htm it says type---one of the type specifiers list, vector, or (vector size), or some other type specifier defined by the implementation to be appropriate.
20:58:16
kpoeck
if its (vector size), why is (defstruct (struct-test-51 (:type (vector 4)))) not legal?
20:58:43
kpoeck
Should that read (vector type) instead of (vector size)?
20:59:11
kpoeck
same text in the franz copy
21:00:13
kpoeck
Ok, is part of the known errata, https://www.cliki.net/ANSI%20Clarifications%20and%20Errata
21:02:03
kpoeck
so it is (vector element-type)
21:03:36
Bike
but in defstruct i wrote cdr instead of cddr so it's broken
Thursday, 12th of December 2019, 23:24:50 UTC