freenode/#sbcl - IRC Chatlog
Search
11:17:29
flip214
I've seen src/code/traceroot.lisp; is there an easy way to get all class instances that point to one thing (or a list of them)?
11:18:03
flip214
Would save me a lot of backward pointers; with quite a few million instances the memory savings might be nice to have.
14:44:37
stassats
just going by the way gc is triggered, it's triggered after allocation, not during
14:45:00
stassats
so if it can't allocate a chunk of memory anywhere it will signal an error during allocation, before GC
17:45:30
flip214
stassats: If there's already a built-in function for going through all live objects looking for a reference, I see that as a valid optimization for rare accesses.
17:57:21
stassats
i would say it would be the obvious mode of operation of any clrhash, not just in sbcl
19:05:49
|3b|
any guess what exit code 2 from make.sh indicates? (running in CI server with no output)
19:24:12
|3b|
and if i modify CI to get logs, "relocation truncated to fit: R_X86_64_32S against symbol `all_threads' defined in .bss section in thread.o" (and also os-common.o)
19:53:04
|3b|
looks like different VM version from the sbcl/sbcl 2.0.8 build , so something could have changed there
19:58:26
|3b|
and apparently i built 2.0.8 on mingw-w64-x86_64-gcc 10.2.0-1 a few days ago with previous vm version
20:01:49
|3b|
would https://github.com/msys2/MINGW-packages/commit/9989b0eceaa5bcff567152910949e6c696406334#diff-2d87c21702e66bae24574c2bd0b8787b affect sbcl?
20:08:06
|3b|
or possibly that is only for building the msys package, and we also need that for 2.0.8
20:15:50
stassats
|3b|: https://github.com/sbcl/sbcl/commit/22844eadcf39f8cbd36850922a21a2450c84c50f
20:39:49
flip214
|3b|: shouldn't really happen... after the initial analysis the data is dumped and reloaded into a new image, so everything visible (at least some known classes) will be live.
20:45:23
|3b|
stassats: i think it is that 10.2.0-3 specifically, not 10 in general, since i built head with 10.1.0 on my machine yesterday, and pretty sure i built 2.0.8 with 10.2.0-1 on CI a few days ago
20:56:36
|3b|
that builds on 10.2.0-1 for me, not sure if i can tell my CI script to pull from git easily or not
21:18:39
|3b|
yeah, built on CI server well enough to fail on some contribs due to leftover env vars from previous tests, so looks good there too