Search
22:49:48
drmeister
compile-file-parallel on linux is not so bad.
22:50:10
drmeister
compiling cmpliteral.lsp serial -> 59 seconds
22:50:19
drmeister
parallel -> 27.6 seconds.
22:50:34
drmeister
kpoeck: Trying "sys:regression-tests;framework.lisp")
22:51:49
drmeister
I'm mystified - I can compile it.
22:51:51
drmeister
https://www.irccloud.com/pastebin/7VR5CfCr/
22:52:03
drmeister
serial 0.722 secs . parallel 0.502 secs
22:52:48
drmeister
This fails ... (load (cmp:compile-file-parallel "sys:regression-tests;framework.lisp"))
22:53:18
drmeister
(load (cmp:compile-file-serial "sys:regression-tests;framework.lisp"))
22:53:41
drmeister
kpoeck: Are you online? Could you tell me exactly what problem you are running into?
22:54:12
Bike
it sounded like compile-file-parallel puts the fasl in the wrong directory?
22:54:27
drmeister
Ok, that I'm investigating.
22:55:48
drmeister
Not the root directory - the ~/Development/clasp (in my case) directory.
22:58:24
drmeister
Ok, I read the log - I missed a bunch of kpoeck's comments.
22:59:04
Bike
also, i won't be in tomorrow or monday... going on a trip
22:59:13
Bike
or probably next thursday since i need to get a root canal
23:09:30
drmeister
Ok - there is a serious problem that pr #918 addresses - but I'd like to change how these problems are fixed.
23:09:33
drmeister
What do I do about this?
23:09:52
Bike
i'm not sure what yo umean
23:10:36
drmeister
https://github.com/clasp-developers/clasp/pull/918/files#diff-d58dcfefc359837b354c57824070cb13R351
23:11:35
drmeister
Shouldn't we move the translate-logical-pathname into write-faso - so that it takes logical pathnames.
23:17:08
Bike
er, but what if it gets a non logical pathname.
23:19:25
drmeister
clhs translate-logical-pathname
23:19:25
specbot
http://www.lispworks.com/reference/HyperSpec/Body/f_tr_log.htm
23:19:56
drmeister
It first coerces the argument to a pathname and then if its a physical pathname it is returned. That's benign - isn't it?
23:20:39
drmeister
kpoeck's solution if a good one - but I don't want write-faso to catch me in the future so I'll move the call into write-faso.
2:24:51
drmeister
There is a recent update to piou - it indicates that it may be working with the latest ASDF.
2:59:25
drmeister
Bike: I hit a cst compiler problem...
2:59:26
drmeister
https://www.irccloud.com/pastebin/ypTHROoP/
2:59:43
drmeister
Quicklisp was loading :exscribe
3:00:32
Bike
do you have a backtrace?
3:01:11
Bike
it doesn't like something in https://github.com/melisgl/named-readtables/blob/master/src/cruft.lisp
3:01:35
drmeister
https://www.irccloud.com/pastebin/kOU6m0KU/
3:02:16
Bike
does (compile nil '(lambda () (progn))) signal an error?
3:02:52
Bike
how about (compile nil '(lambda () (eval-when (:execute))))
3:04:04
Bike
oh, no, it's in our evaluator
3:04:29
Bike
it means we can fix it without messing with cleavir
3:04:43
Bike
https://github.com/clasp-developers/clasp/blob/dev/src/lisp/kernel/cleavir/toplevel.lisp#L123-L129 yeah, i wrote this wrong
3:04:48
Bike
it assumes progns have only one form
3:05:18
Bike
uhh, test by doing (clasp-cleavir::simple-eval-cst (cst:cst-from-expression '(progn)) nil (lambda (&rest ignore)))?
3:05:45
drmeister
https://www.irccloud.com/pastebin/TRhNJ2qM/
3:08:14
Bike
we could instead write that loop as
3:08:16
Bike
(loop with result = nil for cstl = list-cst then (cst:rest cstl) do (setf result (recurse (cst:first cstl) env)) finally (return result))
3:08:25
Bike
i think that would work
3:58:15
drmeister
It doesn't build with that cchange
3:59:06
drmeister
https://www.irccloud.com/pastebin/MxHBY6jb/
3:59:44
Bike
oh, i forgot the end condition
4:00:05
Bike
throw "until (cst:null cstl)" in there i think
4:00:45
Bike
i think? i forget which order things go in
4:01:18
drmeister
https://www.irccloud.com/pastebin/Idj4gXN1/
4:01:53
Bike
yeah. also you can test it without rebuilding using something like the cst-from-expression form i did earlier.
4:05:28
drmeister
Sadly - it doesn't work
4:05:30
drmeister
https://www.irccloud.com/pastebin/8JUPHNfw/
4:06:50
drmeister
Moving the until up one and clearing old code and trying again.
4:07:05
drmeister
Probably should test (null cstl) before the do
5:25:31
beach
Good morning everyone!
5:40:53
drmeister
::notify kpoeck Thank you for that PR - I merged it and made a few additions of my own.
5:40:53
Colleen
drmeister: Got it. I'll let kpoeck know as soon as possible.