libera/#clasp - IRC Chatlog
Search
13:24:06
kpoeck
Hello yitzi I am so exited that we now have CI with regression and ansi-tests, great work!
13:24:07
Colleen
kpoeck: yitzi said 59 minutes, 25 seconds ago: I think this might be ready if you want to try it out https://github.com/clasp-developers/clasp/pull/1310
13:25:27
kpoeck
I think I will now try to integrate cl-bench in koga. I now you can do it quicker than me, but I want to learn
13:26:34
kpoeck
I also have a shell script that tries to compile about 100 quicklisp system and runs their tests. I try that from time to time, just to see whether we have a compilation error
13:27:39
kpoeck
I will copy the script to the wiki, so you can have a look at it and perhaps advise how to best implement this in koga/ninja..
13:29:23
yitzi
Did you see the comment about "draft PR"? Seems like a good way to run the CI tests when a branch isn't ready yet.
13:38:35
kpoeck
I have about 4 pending pr to be uploaded (hopefully not controversial), Perhaps we can wait for a 1.1 tag until these are checked?
13:41:36
yitzi
Sounds good. Just wondering if we should autogenerate other stuff like graphic-char-p also. From the comment in the cc file seems like you have had problems with that also in the past.
15:01:06
Bike
tried the parallel analyzer and i got a relatively straightforward error that "/tmp/process0-output.log" doesn't exist
15:08:01
Bike
(problem two: cl:open should probably not be signaling just a file-error with no description)
15:55:49
mns
is the freebsd branch good to be used, or should I just use 'main' branch and work off that ?
16:14:35
mns
yitzi: is it failing here: boehmprecise/src/llvmo/link_intrinsics.pp /usr/include/sys/elf_generic.h:52:2: error: "Unknown byte order"
16:55:05
yitzi
Ok. The unknown byte error is fixed in the freebsd branch. It is failing in bclasp compilation with some kind of complaint about mutex.
17:38:08
stassats
so i used your github actions to localize the problem, turns out it's caused by trying to call non-existing programs
18:03:01
kpoeck
::notify yitzi here is the promised script to run about 100 quicklisp libraries with clasp: https://github.com/clasp-developers/clasp/wiki/Quicklisp-libraries-in-Clasp . Currently ironclad loops, so I disabled it (is reported as an issues, but i still owe Bike a reproducable test case)
18:04:09
stassats
i tried running automated asdf tests, but asdf:test-system doesn't return anything meaningful?
18:05:28
kpoeck
I believe e.g. `(asdf:test-system (asdf:find-system :djula))` should work if I understood you correctly
18:08:39
kpoeck
Just tried in sbcl, (ql:quickload :djula) and (asdf:test-system (asdf:find-system :djula)) compile the system and run the tests, see https://gist.github.com/kpoeck/0c96a85fb0e6c1129f9b68bc2f33508b for the log.
18:10:56
stassats
(asdf:operate 'asdf:test-op system) _obviously_ doesn't return what the operation returns
18:12:05
stassats
basically have to validate each system at specific revisions and extract their test invocations
18:12:48
stassats
here's the result #<ASDF/LISP-ACTION:TEST-OP > #<ASDF/PLAN:SEQUENTIAL-PLAN {7008D903B3}>
18:21:00
yitzi
The lack of a meaningful result from ASDF:TEST-SYSTEM is mentioned in ASDF manual as I recall.
18:21:00
Colleen
yitzi: kpoeck said 17 minutes, 59 seconds ago: here is the promised script to run about 100 quicklisp libraries with clasp: https://github.com/clasp-developers/clasp/wiki/Quicklisp-libraries-in-Clasp . Currently ironclad loops, so I disabled it (is reported as an issues, but i still owe Bike a reproducable test case)
19:34:01
Bike
parallel analyzer is now giving me "Permission denied" trying to open /tmp/process0-output.log
20:50:34
drmeister
Can you take care of it? It's here... https://github.com/clasp-developers/clasp/blob/main/src/lisp/modules/clasp-analyzer/clasp-analyzer.lisp#L3908
20:51:31
drmeister
This is like the third time where I used a simple /tmp/ filename and we've had this problem. If you have a recommendation for a good standard operating procedure for naming files in /tmp/ I'll adopt it.
21:05:58
Bike
do we even want to touch the filesystem here? i guess so. but if we use an unpredictable name we kind of lose the point of it
21:09:00
yitzi
Shouldn't we be putting that into `build/` thereby avoiding the name clashes and temp files?
21:14:13
yitzi
Doesn't look like I told the static analyzer about the possibility of build directories other that "build/"
21:14:38
yitzi
Bike: If you need to hardwire the path, I can figure out how to make automatic from koga later on.
21:15:23
Bike
i honestly only have a dim understanding of how the analyzer works. my inclination is just to have the base path for these as a keyword argument. i do not know what *compile-commands* is
21:17:12
yitzi
That is fine. If you do a relative path like "foo.log" it will probably just put it into build for you.
0:04:20
drmeister
So - if I add a workflow to cando - could I get that to build the same way? Could I copy your test.yml and modify it?
0:26:09
drmeister
Hmm, that's a question. If I push to the cando repo I'd like to build cando. If we push to the clasp repo - I'd like to build cando. You can't build the cando repo by itself - it's an extension of clasp.
0:29:18
yitzi
Well, the regression tests and the ansi test run with ignore-extensions set. So unless cando has a test suite it would just prove that cando can be built.
0:31:36
yitzi
In any case the short term answer would be to build in clasp's repo and see if we can trigger builds from cando also. Otherwise, I'll have to come up with a different workflow for cando that clones cando into a specific folder inside a clasp clone.
0:35:02
yitzi
Probably clone clasp first, then use actions/checkout@v2 to clone in extensions/cando, then `./koga --extensions=cando --skip-sync=cando`