freenode/#sbcl - IRC Chatlog
Search
10:07:27
Posterdati
leot on #netbsd applied a patch to sbcl source three to bootstrap and install on netbsd amd64, I applied that patch on the git version of sbcl. It is possible to have it in the official release?
10:26:27
leot
If Posterdati has not shared that here the patch that fixes sbcl on NetBSD 8.0 and newer:
10:26:31
leot
http://cvsweb.netbsd.org/bsdweb.cgi/~checkout~/pkgsrc/lang/sbcl/patches/patch-src_runtime_GNUmakefile?rev=1.1&only_with_tag=MAIN
10:28:06
leot
I think that something similar can be handy also for Linux, that always have a paxctl program but the syntax is a little different
10:29:04
leot
Regarding OpenBSD, they do not have any ELF annotation and needs to adjust that by passing flags to the linker, so anything like that will probably not relevant for them (and from a quick look to the OpenBSD port sbcl they already do that)
10:42:20
scymtym
Posterdati: sorry, i was doing something else. #sbcl is often slow to respond but you often get an answer eventually. in this case, i'm not yet sure whether i will be able to help, but somebody else may pick it up later
10:46:03
scymtym
Posterdati: i didn't get what should be added the table (the platform table on sbcl.org, i assume?)
10:51:27
leot
scymtym: Oh, and forgot to say, this is how it is defined as part of the build phase on NetBSD `SBCL_PAXCTL="paxctl +m"'
10:52:26
leot
...and I guess that with paxctl of PaX team the corresponding incantation will be `paxctl -m' (just by skimming the man page!)
10:56:09
leot
...so, with that patch I guess that instead patching that on the fly they can just SBCL_PAXCTL="pax-mark -mr" and that will do the trick
11:22:34
scymtym
Posterdati: could you explain again what change you were suggesting regarding the platform table?
11:28:27
scymtym
leot: does this look ok to you? https://github.com/scymtym/sbcl/commit/5d1af46703344629a10ede3330402bcb12f60962
12:44:56
scymtym
Posterdati: ok, i see what you mean. i don't think we have a way of indicating that a particular platform+version combination works without linking to binaries. as you probably have noticed, all green cells link to binary downloads
13:24:06
scymtym
a separate table. i don't think we would manage to keep that up to date. but maybe somebody else thinks otherwise
13:25:15
scymtym
i'm not saying the information wouldn't be useful. i'm just doubtful regarding the practicality
14:19:23
Xach
Is there a way to find out the full pathname of a shared object that appears as relative in sb-sys::*shared-objects*?
14:33:42
stassats
a new SET kind or a lexenv declaration, the latter could be used for something else, but how do i name it
14:45:43
Xof
does anyone have capacity / interest to follow up? It might be nice to be paid for some sbcl work for someone. (I am about 180% employed at the moment; it's not me for the forseeable future)
14:48:29
stassats
none of the stylewarning's attempts to woo me into working on PPC has succeeded so far
14:49:22
stassats
(defun f4 (x) (setf x 30) (catch 'tag (setf x 20)) x) and here i'm incorrectly removing (setf x 30)
14:51:57
pfdietz
I read that on ARM64 there's a mode where the top 8 bits of addresses are ignored. So many tags.
14:53:51
stassats
i think only CATCH matters right now, any other NLX go through local functions, which i ignore atm
15:18:18
stassats
i think i'll postpone set/let elimination till the next release, as there is a lot of corner cases
15:18:39
pfdietz
I wonder if this sort of optimization could be useful in SBCL. https://arxiv.org/pdf/1807.06735.pdf