libera/#sicl - IRC Chatlog
Search
14:40:03
pfd
It's my sbcl. It wasn't behaving consistently at the command line. Sometimes upon invocation it produced strange errors. So, that must be the culprit.
20:21:47
pfd
So, each time I retried with a different, earlier version of SBCL, down to 2.1.9,, each time deleting and recloning my SICL folder,along with clearing out my .cache and .sbclrc file.
20:21:48
pfd
I've tred 3 or 4 times with each version of 3 versions sbcl, and I keep running into aprrox 23 fatal errors after about the 42000th line. All these errors are happening between the 41,900th line and about the 44,000th line.
20:23:26
pfd
Also twice I uninstalled by delecting the depencies from the quicklisp/local-projects folder. The only thing I haven't tried is to delete the QL folder entirely; but I don't see how that might help.
21:19:29
Bike
i tried it just now and rather than a bunch of indetermine errors, i just get an error that sicl-environment depends on stealth-mixin, which doesn't exist. so that should probably be added to the readme
21:22:27
Bike
after that i hit some problems because the repo cleavir is not compatible with the one in sicl, but if i remove the repo cleavir that goes away and i can load sicl-boot fine
21:25:40
Bike
when actually bootstrapping, it gets most of the way through, but then fails because it can't find cyclosis-intrinsic
21:26:59
Bike
it's kind of difficult to do any specific troubleshooting based only on you hitting a bunch of fatal errors, and not like, what those errors are
21:27:16
Bike
even though asdf does its stupid interception thing, you should be getting the actual compiler conditions dumped to stdout
21:30:23
Bike
i would double check that sicl isn't trying to load the repo cleavir (i.e. the one clasp uses), because they are unfortunately incompatible right now
21:32:37
Bike
i don't think the sbcl version is the culprit. sicl is not doing anything implementation specific if it can help it
21:50:58
pfd
I installed stealth-mixin from QL and I also had to re-install mcclim because I decided to nix my whole QL folder and start over. I had some details in there which were write-protected; which didn't make sense to me.
21:52:56
pfd
After that I still have tens of fatal erorrs by the time I finish stage 1. The main, repeating error is that the lisp files are not getting compled to .fasl files and the process keeps complaining about missing .fasl files
22:05:29
Bike
that's not an informative error. that's asdf just saying it had problems while compiling. if you look at the output outside of the debugger, you'll probably see some actual errors.
22:06:00
Bike
the clasp install process does download its dependencies into directories internal to it, and shouldn't interfere with anything sicl is doing. i'm just spitballing
22:13:25
pfd
0: (UIOP/LISP-BUILD:CHECK-LISP-COMPILE-RESULTS NIL T T "~/asdf-action::format-action/" ((#<ASDF/LISP-ACTION:COMPILE-OP > . #<ASDF/LISP-ACTION:CL-SOURCE-FILE "sicl-boot-base" "utilities">)))
22:13:26
pfd
CONTEXT-ARGUMENTS = ((#<ASDF/LISP-ACTION:COMPILE-OP > . #<ASDF/LISP-ACTION:CL-SOURCE-FILE "sicl-boot-base" "utilities">))
22:16:05
pfd
; Stream: #<SB-INT:FORM-TRACKING-STREAM for "file /home/pfd/common-lisp/SICL/Code/Boot/utilities.lisp" {1006C1C133}>
22:16:06
pfd
; compiling file "/home/pfd/common-lisp/SICL/Code/Boot/utilities.lisp" (written 14 MAY 2022 05:25:31 PM):
22:47:30
Bike
https://github.com/s-expressionists/Eclector/commit/c6c76ab4f57db9a812d107e21251b65ff69078d5 looks like *client* being in that package is a year old
22:51:37
pfd
As mentioned earlier, I completely deleted my quicklisp folder and when I re-cloned SICL, a little earlier I simply ran Robert's get-depencies.sh file, which I hope pulls the latest version of the needed dependencies.
22:53:50
pfd
Also, there are differences between the SBCL versions. This particulare Line15 double error only occurs with SBCL 2.2.4. I got down to lines 41000 with 2.2.3, 2.2.2 and 2.1.9 . Those three version presented fatal 23 fatal errors,
22:54:52
pfd
Mind you this was a cursory observation. If it helps Robert, I can do a meticulous and careful step by step account with each SBCL version, etc.
23:58:07
Bike
did you nuke cached fasls? they can sometimes screw things up. should be in ~/.cache/common-lisp/
23:58:38
Bike
i really think the sbcl version differences are spurious. nothing in sicl is depending on bleeding edge sbcl
23:59:05
Bike
you can also do (ql:where-is-system :eclector) to make sure you're loading the right thing
0:21:37
pfd
Why the heck does it depend on being in my ~/.cache/common-lisp/? Isn't QL smart enough to put it there on first invocation?
0:21:38
pfd
Anyhow, I'll try ql:where-is-system :eclector, but I really don't think that's the problem. Let me check, one sec.
0:29:13
pfd
:Bike My copy of Eclector is exactly where it's supposed to be; as placed there by Robert's script, in '~/quicklisp/local-projects/Eclector'.
0:30:42
pfd
After dinner I'm going to nix SBCL 2.24 and downgrade to SBCL 2.2.2. If I'm right there's something different about 2.2.4.
1:24:53
Bike
pfd: does the eclector downloaded by the script match the eclector on github? because the eclector on github definitely has *CLIENT* in ECLECTOR.BASE
1:31:50
pfd
Yes, I understand that the fasls go into the .cache. So that's where the compiler should put them before the rest of the bootstrapping needs them.
1:31:51
pfd
Robert's script shows that it's pulling Eclector from the correct github s-expressionts account. So let me take a peek at this ECLECTOR.BASE file in my copy. One sec.
1:36:21
Bike
that is where the compiler put them. what i am referring to is that sometimes, if you compile one version of a source, asdf will cache the fasls there. then it turns out you were using the wrong source, but the fasls are still there, so when you try to load the new source you get the old fasls.
1:37:22
pfd
That's why I was clearing my .cache after every failed attempt. I was doing this methodically.
1:38:26
Bike
https://github.com/s-expressionists/Eclector/tree/master/code/base this is what files should be there, and if you look at package.lisp, you should see an export of *client*
1:43:13
Bike
okay. so that's the right file, probably. and probably _not_ the file that was being loaded earlier when you saw that error referring to *client*.
1:43:30
Bike
and yes, please use some kind of paste service for anything more than a line or so. any of them should do
1:44:24
Bike
i don't know what to tell you. that package definition very obviously exports the symbol *CLIENT* from ECLECTOR.BASE. the error message said there was no such symbol in that package.
1:44:26
pfd
So, give me some time to methodically experiment as per my suspicion. I'll be open minded towards being completely wrong.
1:46:27
pfd
BTW what version of SBCL were you able to successfuly bootstrap this latest iteration of SICL.
1:47:46
pfd
Thanks. I'll step back to that version, which I recall didn't hiccup at line 15! I'll also check SBCL 2.2.3.