libera/#clasp - IRC Chatlog
Search
11:47:17
drmeister
My demo last night to the class went sideways a bit when we lost contact with Ohio.
12:20:08
drmeister
Yes - either that or that machine was still not powerful enough for 16 users. They were working for a while but then it just lost contact although the machine reported as being up.
12:23:13
drmeister
All together it took me about an hour to bring the whole thing up but I got caught up for about two hours trying to set the AMBERHOME environment variable for each user.
12:23:47
drmeister
I still have a problem with libc++ not being found on one of my people's Macbook Air's.
12:24:09
drmeister
I'm going to get that machine in my hands in about 15 min and then I'm going to compare it to another machine.
12:32:10
drmeister
yitzi: The cando-user-install command - it creates the link command that looks like:
12:32:33
drmeister
`/usr/local/Cellar/llvm@14/14.0.6/bin/clang++ -L/usr/local/Cellar/llvm@14/14.0.6/lib -Wl,-search_paths_first -Wl,-headerpad_max_install_names -pthread -fvisibility=default -rdynamic -L/usr/local/Cellar/expat/2.4.8/lib -L/usr/local/Cellar/fmt/9.1.0/lib -L/usr/local/Cellar/gmp/6.2.1_1/lib -o/Users/meister/.local/bin/scando -sectcreate __CLASP __clasp /tmp/ss-Ey9pkHfZ
12:32:33
drmeister
-Wl,-force_load,/usr/local/Cellar/cando/1.0.0-529-gc6c8ae3c6-g7371ab99_2/lib/clasp//libclasp.a -lclang-cpp -lLLVM-14 -lexpat -lfmt -lgmpxx -lgmp`
12:33:15
drmeister
I'd like to run that command over and over again to debug this problem. I need the /tmp/ss-Ey9pkHfZ temporary file to not disappear.
13:15:47
yitzi
drmeister: The link command is created by clasp techically. Its buried in the snapshot C++ code.
13:33:37
Bike
this is so weird. i thought nuking .cache and .slime/fasl would be enough but apparently not.
13:34:04
Bike
yeah, i checked there, i don't see any fasls... there's bitcode, and there was a faso for sockets, but not asdf
13:35:15
yitzi
If we could get this issue fixed with the analyzer then I build the modules properlly
13:35:47
Bike
yeah, i did that too... i don't _think_ there's anything there, but let me check again
14:27:20
yitzi
sysroot is .... /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/
14:28:03
yitzi
the option is -isysroot /Applications/Xcode.app/Contents/Developer/Platforms/MacOSX.platform/Developer/SDKs/MacOSX.sdk/
14:42:03
drmeister
Keep that in mind - we could cut out the iclasp-boehm build step. Just load all of the cclasp code and then go to the static analyzer. Run the whole thing in bytecode.
14:56:33
drmeister
The gdb frame filters and frame descriptors thing would let me change the trampoline frames so they display lisp names and arguments.
15:24:43
drmeister
Bike: You have 'vm' with the interpreter and aclasp/bclasp compiler completely disabled - right?
15:26:33
drmeister
::notify yitzi (Q1) Do you use quicklisp to load the static analyzer? (Q2) Can we merge 'vm' into 'yitzchak_vm-lpt'?
15:41:48
Colleen
yitzi: drmeister said 15 minutes, 15 seconds ago: (Q1) Do you use quicklisp to load the static analyzer? (Q2) Can we merge 'vm' into 'yitzchak_vm-lpt'?
15:43:34
yitzi
rebase is probably better. Or we just merge yitzchak_vm-lpt and commit to fixed the static analyzer now.
15:45:11
yitzi
But it doesn't really matter. You can make changes to that branch if you want. You have the bridge.
16:03:45
drmeister
yitzi: Can you fix the CLASP_STAGE COUNT thing - it looks like clasp-builder.lisp isn't listening to it.
16:05:10
drmeister
Here's what I'm thinking. I will make changes in the 'vm' branch to improve debugging with the vm. Better backtraces mainly.
16:05:28
drmeister
Then I'll merge that into yitzchak_vm-lpt and use it to fix the static analyzer problem.
16:08:47
drmeister
Actually - forget the bclasp thing - I'll load_vclasp-boehm and then run from that.
16:09:04
drmeister
It doesn't take any longer and it will give me the option of compiling some stuff to native code.
16:21:22
Bike
now most of the stuff to remove is environment classes and lambda list handlers, but i'll need the analyzer for that, i think
16:47:39
yitzi
They are defined in foundation.lisp ... might be nice to have in clasp-builder, though. Trying to define and see if that works.
17:13:05
drmeister
yitzi: I'm doing `export CLASP_STAGE_COUNT=3` and then `ninja -C build load_cclasp-boehm`
17:25:00
yitzi
drmeister: try adding `(defun 1- (num) (- num 1)) (defun 1+ (num) (+ num 1))` to init.lisp
17:31:02
drmeister
It's when it tries to get the (file-write-date source-path) in clasp-builder.lisp:581
17:36:54
drmeister
You know what - I don't want to stop there. I had a goofy idea that we could run the static analyzer with just the bytecode compiler and I thought stopping at an earlier stage would be useful.
17:37:31
drmeister
What I do want to do is rebase or merge 'vm' into 'yitzchak_vm-lpt' - how do you recommend that I do that?
17:37:45
yitzi
I think that is a good idea, but I would feel better if we could fix one thing at a time. :)
17:39:01
yitzi
If we can get the static analyzer to work on clasp (I can work on cando later) we can merge yitzchak_vm-lpt back into vm
17:39:38
drmeister
I have improvements in the python debugging extension in vm that would be helpful here.
17:41:31
yitzi
I am working on reusing the cclasp faso files in the eclasp image. That will make the build even faster.
18:00:27
drmeister
I think the issue is that the system (list of files) should be limited to just what was loaded.
18:02:22
drmeister
I can iterate faster on debugging it I run the static analyzer only with bytecode.
18:07:42
yitzi
If you want to you can try moving #@"cclasp-translations.lisp" from the start of stage 5 to the end of stage 4
18:09:06
drmeister
Ok, I set it up so that load-stage returns the list of system entries that it actually loaded, in the order they were loaded.
18:13:33
yitzi
Ok....I am confused. You can't change the return value of load-clasp....compile-clasp depends on that.
18:13:47
drmeister
I see - I may not have needed to do what I did if I'd move the cclasp-translations.lisp to the end of stage 4
18:14:47
drmeister
It won't change the return value of load-clasp unless you CLASP_STAGE_COUNT = <something less than the last stage>
18:16:12
yitzi
I'd have to see the code. Plus, I don't really understand what you are trying to accomplish.
18:28:28
drmeister
Bike: I want to create a 'vm-boehm' branch that merges 'vm', 'yitzchak_vm-lpt' and your changes that eliminate the interpreter and aclasp/bclasp compiler.
18:33:43
Bike
activation frames, environments, and lambda list handlers are still on the chopping block
18:45:28
drmeister
It went and created a pull request? I don't want a pull request - I want a branch!?