Search
Sunday, 25th of September 2022, 11:16:37 UTC
15:46:58
drmeister
I'm working on the static analyzer using cclasp - something is up with unwinding and the bytecode compiler.
22:18:40
drmeister
'vm' branch isn't building
22:18:42
drmeister
https://usercontent.irccloud-cdn.com/file/CksUgs46/image.png
22:21:52
yitzi
Is that vclasp or the old cclasp?
22:22:24
drmeister
right - I should be building vclasp?
22:23:10
yitzi
The bytecode version is called vclasp on vm. cclasp on vm-boehm
22:23:27
drmeister
I think I forgot that.
22:23:43
drmeister
I've wasted most of the weekend trying to figure out what is going on with vm-boehm.
22:23:55
yitzi
If cclasp doesn't build on vm, then the static analyzer probably won't either.
22:23:57
drmeister
I'm going back to vm to get a reference build that works to compare it to.
22:24:46
drmeister
Great - then we are way out on a limb until we figure out what is effed up.
22:24:47
yitzi
I don't think the breakage on vm-boehm has anything to do with Bike's changes.
22:24:58
drmeister
I have no idea right now.
22:25:08
yitzi
yitzchak_vm-lpt's static analyzer was broken also.
22:25:16
yitzi
And that didn't have Bike's changes.
22:26:43
drmeister
This is what I get trying to build vclasp-boehmprecise on 'vm'
22:26:50
drmeister
https://www.irccloud.com/pastebin/UUG11FM9/
22:27:44
drmeister
I really don't like CLASP_STAGE_COUNT
22:28:24
yitzi
Sounds like the load-time is slightly broken there
22:28:44
drmeister
No, I had CLASP_STAGE_COUNT
22:29:16
drmeister
No, I had CLASP_STAGE_COUNT=5
22:32:02
yitzi
For me the static analyzer error starts with an error triggered by a non-existent file inside OPEN. I suspect that ASDF intends to handle the error, but it just keeps unwinding.
22:34:09
drmeister
Yes, that's what I'm getting.
22:34:34
drmeister
It's in an ignore-error and it's not ignoring it.
22:34:41
yitzi
Maybe we need a simple reproducer.
22:37:50
yitzi
The idea that someone would use IGNORE-ERRORS on OPEN with it configured to return an error on a non-existent file is kinda dumb.
22:38:20
drmeister
That looks like what they did though.
22:39:00
yitzi
Do you know where in ASDF it is coming from?
22:41:56
drmeister
(defun process-source-registry-cache (directory collect)
22:42:06
drmeister
If its the same problem.
22:42:23
drmeister
I'm building clasp now so that I can trace them both and see if I can spot a difference.
22:51:30
yitzi
If ASDF is that carefree with allowing conditions to be signaled and then quashing them that might explain why Clasp quicklisp loading was slow in the past.
23:11:03
drmeister
Yes, it is that carefree with allowing conditions to be signalled...
23:11:15
drmeister
https://usercontent.irccloud-cdn.com/file/wXPx3OF8/image.png
23:11:36
drmeister
Note how the inner functions don't return.
23:14:14
drmeister
https://usercontent.irccloud-cdn.com/file/uzooDkTf/image.png
Sunday, 25th of September 2022, 23:16:37 UTC