freenode/#lisp - IRC Chatlog
Search
10:43:08
pjb
Indeed, sbcl --noinform --no-userinit --eval '(progn (print (find-package "SB-POSIX")) (terpri))' --quit # prints NIL
10:44:33
pjb
try: sbcl --noinform --no-userinit --eval '(progn (mapcar (function print) (sort (mapcar (function package-name) (list-all-packages)) (function string<))) (terpri))' --quit
10:45:13
pjb
and try: sbcl --noinform --no-userinit --eval '(require "SB-POSIX")' --eval '(progn (print (find-package "SB-POSIX")) (terpri))' --quit
12:35:17
Harag
I can require sb-posix and chdir to root but (user-homedir-pathname) is still /home ...
12:37:20
Harag
and it looks like (asdf:oos 'asdf:load-op "quicklisp" :verbose nil) is failing with Component "quicklisp" not found in quicklisp/setup.lisp
12:37:30
Harag
which leads to ((:METHOD ASDF/OPERATE:OPERATE (SYMBOL T)) ASDF/LISP-ACTION:LOAD-OP "quicklisp" :VERBOSE NIL) [fast-method]
12:38:47
Harag
which leads to ((:METHOD ASDF/OPERATE:OPERATE (SYMBOL T)) ASDF/LISP-ACTION:LOAD-OP "quicklisp" :VERBOSE NIL) [fast-method] changed too much to cope with.
12:41:42
Harag
so I have put a lot of prints in it to see what is happening and am loading it with --load and not .sbclrc
12:43:39
jackdaniel
maybe try to "install" quicklisp at a location where paths do not change (changing paths sounds like an awful way to do things, probably there is a good reason for that though)
12:45:44
Harag
yeah I am not the one changing the HOME from /root to /home "google cloud run" CMD is for some reason. So I installed ql in a custom folder /src/ and pointed init to that but will try to install it in /home
12:48:28
Harag
while its compiling... I do the install with RUN in the docker file and run reports the HOME to be /root but when you use CMD it reports HOME as /home ...
12:53:01
Harag
the RUN command has no issues with QL but the CMD has and the only diff I could see was the HOME /root vs /home
12:55:07
jackdaniel
(so, due to impossibility to write, you signal another error, which can't be written)
12:55:58
Harag
which leads me back to trying to debug quicklisp/setup.lisp to see what caused the first error...
12:57:22
jackdaniel
the first error may be for the same reason, i.e that quicklisp can't write something like "may the force be with you"
12:59:36
Harag
can you take a guess at what linux command the CMD is doing to change the HOME but not the actual user, I checked the user is 0 in RUN and CMD and I cant see that they are using su or something to change users ps reports all processes to be using root
13:00:44
jackdaniel
I've decided long time ago to not pay attention to systemd and docker hacks in the kernel, life is too short to play catch with kludges
13:03:13
jackdaniel
I remember that sbcl's mutexes on google vms were not fair, what was a major disaster (not to mention amount of time the person who tracked that have wasted)
13:05:34
Harag
because you have to do some really crazy stuff the connect to it...basically from what I found on google you have to setup a machine with a fixed ip so that the docker can do a call back and connect to you and then you can work with it
13:05:39
jackdaniel
you will probably need to change the loopback interface and that will not be safe, but for debugging that should be good enough
13:40:02
Harag
well its the (asdf:oos 'asdf:load-op "quicklisp" :verbose t ) that is causing the "11 nested errors " but cant get more info out of it than that...
13:48:16
pjb
try: sbcl --noinform --no-userinit --eval '(load #P"~/quicklisp/setup.lisp")' --eval '(ql:quickload :split-sequence)' --eval '(progn (print (split-sequence:split-sequence (quote /) (quote (a / b / c)))) (terpri))' --quit
14:10:39
Harag
now I just need to put it in the back ground, just running woo blocks the deploy from ever completing
14:26:46
Harag
I suspect all the trouble is stdin, "google cloud run" does not allow you to pass parameters to fake it like you can in docker native with -ti
14:28:51
bitmapper
Xach: i noticed your twitter post, how much have you worked with the Obj-C interface in CCL?
14:35:21
bitmapper
i was working on updating the IDE to add dark mode support a bit ago and it was quite nice
14:36:28
bitmapper
Xach: the only thing is the interfaces are quite old, and they are not done updating the software that generates them
16:21:21
Josh_2
I want to sleep N amount of times in 1 second, how do I do this? Say I want to sleep 5000 times, how do I get the number I pass to sleep?
17:05:57
pve
Hi, I recently saw a library that allowed for better error reporting in macros, but I can't remember the name. Does this ring a bell for anyone?
18:57:09
heisig
theseb: I once wanted to write a compiler in Clojure. Turns out it takes 6 seconds to even start the process. So we ditched it for CL. No regrets so far :)
18:57:24
pjb
theseb: there are two CL libraries implementing Clojure. If you have fun with that, then have fun!
18:58:43
heisig
I think the main selling point is JVM integration and its immutable data structures. Sounds like we should advertise ABCL a bit more.
19:04:55
theseb
pjb: one thing i really like is they did away with defining lists in terms of conses....instead they define a "sequence" type from which lists and vectors are built
19:05:10
heisig
theseb: That was a few years ago. But I just checked again, 'time clojure -e "(+ 2 3)"' takes 2.2 seconds on my machine.
19:08:09
theseb
heisig: well i'm guessing that just startup...still...not sure why so slow to start up
19:12:13
heisig
theseb: Yes, that's mostly JVM startup time. But it also makes it quite useless for applications that are invoked frequently.
19:13:39
pjb
heisig: well, obviously, you wouldn't keep booting them all the time. You'd do like Android, you'd boot once and that's it.
19:46:19
theseb
i know why "or" must be a macro....but does "and" have to be a macro? I can't remember why I implement "and" as a macro
19:46:35
theseb
The only reason I can see is that if the 1st arg is false...you should NOT eval the 2nd arg
20:34:41
phoe
beach: reddit made me aware of http://blog.rongarret.info/2010/02/new-and-improved-lexicons-now-50-lexier.html - are you aware of that work?