2016-08-11T00:00:32Z stassats: how often do you redefine such functions?
2016-08-11T00:02:59Z scymtym: doesn't each compile-file+load even without redefinition currently generate a new ctor? i can image having hundreds of those in a single file and recompiling+loading the over repeatedly
2016-08-11T00:03:20Z scymtym: something similar led to the discovery of the problem
2016-08-11T00:03:31Z stassats: if you both redefine and call those functions
2016-08-11T00:03:35Z stassats: how often does that happen?
2016-08-11T00:04:04Z stassats: anyway, the compiler-macro->source-transform change is good either way, i'm making it
2016-08-11T00:05:01Z scymtym: maybe i misunderstood. i thought ctors for constant class names were created at load-time while inline caches (which only grow at call-time) were used for non-constant class names
2016-08-11T00:05:38Z stassats: oh right, there's load-time-value, but my point still stands
2016-08-11T00:05:44Z stassats: it's a rare situation
2016-08-11T00:08:52Z scymtym: i'm not sure i agree that it's rare. but even if it's rare, shouldn't repeated compile-file+load of an unmodified source file be mostly idempotent?
2016-08-11T00:09:08Z stassats: a fasl from (defvar *x* (make-instance 'foo)), was 1860 bytes, 614 now
2016-08-11T00:09:15Z scymtym: nice
2016-08-11T00:19:36Z stassats: what we need is a way to gc away ctors
2016-08-11T00:20:33Z scymtym: yes, that would be best
2016-08-11T00:21:09Z stassats: that would need what, *all-ctors* to be weak?
2016-08-11T00:21:42Z stassats: and the class plist
2016-08-11T00:23:30Z stassats: got a bunch of ctor.impure.lisp failures..
2016-08-11T00:26:52Z stassats: oh, it does (setf (compiler-macro-function 'make-instance) wrapper)
2016-08-11T00:28:53Z stassats: and there i though something was really wrong
2016-08-11T00:30:08Z stassats: ok, pushed the transform thing, will try to play with weak pointers to ctors sometime in the future
2016-08-11T00:31:32Z stassats: it would need a finalizers, wouldn't it? and finalizers are piss poor in sbcl
2016-08-11T00:32:06Z scymtym: we also had to stop putting constructors in globaldb
2016-08-11T00:32:55Z stassats: so, what, make *all-ctors* into a weak ht?
2016-08-11T00:33:09Z stassats: and weak pointers in the class plist
2016-08-11T00:33:17Z stassats: sounds like a plan
2016-08-11T00:35:25Z oleo quit (Read error: Connection reset by peer)
2016-08-11T00:35:40Z stassats: wouldn't need those without-package-locks
2016-08-11T00:36:02Z oleo joined #sbcl
2016-08-11T00:36:05Z scymtym: worth a try, i think. i'm out of time for today, though
2016-08-11T00:37:12Z stassats: and a word lock can go, just a locked ht
2016-08-11T01:21:17Z em1l_ joined #sbcl
2016-08-11T01:24:58Z em1l quit (Ping timeout: 265 seconds)
2016-08-11T01:42:15Z vydd quit (Read error: Connection reset by peer)
2016-08-11T01:47:36Z stassats: another question, do we really need to invoke compiler macros for fop compilation?
2016-08-11T01:49:05Z vydd joined #sbcl
2016-08-11T02:17:00Z stassats quit (Ping timeout: 276 seconds)
2016-08-11T02:54:31Z vydd_ joined #sbcl
2016-08-11T02:56:22Z vydd quit (Ping timeout: 244 seconds)
2016-08-11T03:30:57Z shka_ joined #sbcl
2016-08-11T03:46:49Z akkad quit (Ping timeout: 250 seconds)
2016-08-11T03:48:07Z mgodshall quit (Ping timeout: 250 seconds)
2016-08-11T03:48:15Z mgodshall joined #sbcl
2016-08-11T03:56:29Z akkad joined #sbcl
2016-08-11T05:21:00Z oleo quit (Quit: Leaving)
2016-08-11T05:36:33Z nyef quit (Ping timeout: 240 seconds)
2016-08-11T05:41:50Z dsp- joined #sbcl
2016-08-11T05:43:16Z DGASAU` joined #sbcl
2016-08-11T05:48:39Z dsp-_ quit (Ping timeout: 250 seconds)
2016-08-11T05:48:39Z DGASAU quit (Ping timeout: 250 seconds)
2016-08-11T06:12:41Z shka_ quit (Ping timeout: 244 seconds)
2016-08-11T07:01:47Z vydd_ quit (Remote host closed the connection)
2016-08-11T07:01:52Z DeadTrickster joined #sbcl
2016-08-11T07:03:25Z DeadTrickster_ joined #sbcl
2016-08-11T07:29:50Z nzambe quit (Quit: http://www.kiwiirc.com/ - A hand crafted IRC client)
2016-08-11T07:31:12Z vydd joined #sbcl
2016-08-11T07:42:46Z gingerale joined #sbcl
2016-08-11T07:48:07Z angavrilov joined #sbcl
2016-08-11T08:26:40Z Bike quit (Quit: slepd)
2016-08-11T08:26:50Z sjl__ joined #sbcl
2016-08-11T08:27:27Z sjl quit (Ping timeout: 264 seconds)
2016-08-11T08:52:11Z barbone quit (Remote host closed the connection)
2016-08-11T09:02:28Z angavrilov quit (Quit: Konversation terminated!)
2016-08-11T09:02:40Z angavrilov joined #sbcl
2016-08-11T09:05:38Z barbone joined #sbcl
2016-08-11T09:20:07Z barbone quit (Quit: KVIrc 4.2.0 Equilibrium http://www.kvirc.net/)
2016-08-11T09:32:59Z barbone joined #sbcl
2016-08-11T09:57:56Z barbone quit (Remote host closed the connection)
2016-08-11T10:10:54Z barbone joined #sbcl
2016-08-11T10:23:57Z attila_lendvai joined #sbcl
2016-08-11T10:53:24Z attila_lendvai quit (Ping timeout: 260 seconds)
2016-08-11T10:55:53Z christoph_debian: hm people
2016-08-11T10:56:05Z christoph_debian: I have a old patch for concurrencyon powerpc
2016-08-11T10:56:09Z christoph_debian: from around 1.2.4
2016-08-11T10:56:21Z christoph_debian: I guess it's fixed properly these days?
2016-08-11T10:56:43Z christoph_debian: https://pbot.rmdir.de/U6uHn-Og3DN-wEHvgimygQ
2016-08-11T10:58:30Z karswell quit (Remote host closed the connection)
2016-08-11T10:58:54Z karswell joined #sbcl
2016-08-11T10:58:55Z barbone quit (Ping timeout: 244 seconds)
2016-08-11T11:07:02Z barbone joined #sbcl
2016-08-11T11:10:13Z karswell quit (Ping timeout: 258 seconds)
2016-08-11T11:41:41Z DGASAU` is now known as DGASAU
2016-08-11T11:54:03Z DeadTrickster_ quit (Ping timeout: 244 seconds)
2016-08-11T11:55:02Z stassats joined #sbcl
2016-08-11T12:41:09Z blackwolf quit (Quit: ERC (IRC client for Emacs 24.5.1))
2016-08-11T12:43:49Z DeadTrickster quit (Ping timeout: 244 seconds)
2016-08-11T12:45:06Z barbone quit (Remote host closed the connection)
2016-08-11T13:18:07Z DeadTrickster joined #sbcl
2016-08-11T13:28:14Z DeadTrickster quit (Ping timeout: 244 seconds)
2016-08-11T13:32:19Z nyef joined #sbcl
2016-08-11T13:46:46Z DeadTrickster joined #sbcl
2016-08-11T13:53:56Z pjkjqu joined #sbcl
2016-08-11T13:56:08Z oleo joined #sbcl
2016-08-11T13:56:08Z oleo quit (Changing host)
2016-08-11T13:56:08Z oleo joined #sbcl
2016-08-11T13:57:01Z miscanthus quit (Ping timeout: 258 seconds)
2016-08-11T14:22:28Z sjl__ is now known as sjl
2016-08-11T14:37:19Z vydd quit (Remote host closed the connection)
2016-08-11T14:40:45Z barbone joined #sbcl
2016-08-11T14:50:19Z nzambe joined #sbcl
2016-08-11T14:57:24Z sjl quit (Read error: Connection reset by peer)
2016-08-11T15:09:05Z dougk_ quit (Ping timeout: 258 seconds)
2016-08-11T15:09:51Z nyef quit (Ping timeout: 276 seconds)
2016-08-11T15:22:00Z dougk_ joined #sbcl
2016-08-11T15:47:36Z sjl joined #sbcl
2016-08-11T15:59:53Z fiddlerwoaroof quit (Quit: Gone.)
2016-08-11T16:01:27Z shka_ joined #sbcl
2016-08-11T16:04:54Z dougk_ quit (Ping timeout: 250 seconds)
2016-08-11T16:17:30Z dougk_ joined #sbcl
2016-08-11T16:19:08Z DeadTrickster quit (Ping timeout: 265 seconds)
2016-08-11T16:54:32Z sjl quit (Ping timeout: 240 seconds)
2016-08-11T17:39:37Z Bike joined #sbcl
2016-08-11T17:39:53Z sjl joined #sbcl
2016-08-11T17:52:09Z nyef joined #sbcl
2016-08-11T18:16:12Z vydd joined #sbcl
2016-08-11T18:34:16Z DeadTrickster joined #sbcl
2016-08-11T19:53:39Z stassats: scymtym: weakining ctors seems to be working well, running tests and pushing
2016-08-11T20:40:41Z Carisius joined #sbcl
2016-08-11T20:46:21Z stassats: scymtym: and it's in
2016-08-11T20:59:30Z gingerale quit (Remote host closed the connection)
2016-08-11T21:07:10Z scymtym: stassats: that's great. i didn't have any time for sbcl today
2016-08-11T21:07:34Z stassats: i also specialized NIL, it's no longer treated as a list
2016-08-11T21:09:35Z scymtym: yeah, i did this as well in my smaller-scoped fix (not committed)
2016-08-11T21:09:38Z shka_ quit (Ping timeout: 244 seconds)
2016-08-11T21:11:24Z scymtym: looking at your commit, didn't we have double earmuffs for DEFGLOBALs as a convention or am i misremembering?
2016-08-11T21:13:00Z stassats: i don't think it's particularly important
2016-08-11T21:13:29Z scymtym: sure, i'm half-asking, half-remarking
2016-08-11T21:14:07Z stassats: there's also %%
2016-08-11T21:14:10Z scymtym: heh
2016-08-11T21:15:01Z scymtym: i'm building now to try the esrap use case
2016-08-11T21:15:01Z stassats: if it were a new variable, i would have given it more thought
2016-08-11T21:16:49Z stassats: can make/allocate-instance return multiple values?
2016-08-11T21:17:40Z stassats: i suppose a cmov load doesn't matter much compared to allocation
2016-08-11T21:18:57Z scymtym: there may be some trick with multiple values when allocating conditions, iirc
2016-08-11T21:21:08Z stassats: returning one value is 4 bytes shorter and very slightly faster
2016-08-11T21:21:47Z stassats: well, i suppose we don't check what make-instance actually returns, so blindly trusting it to return one value may blow the stack
2016-08-11T21:22:45Z stassats: verdict, it's not worth it
2016-08-11T21:22:49Z scymtym: the repeated COMPILE-FILE that got slower by 2 seconds for each repetition is now constant at around 80 ms
2016-08-11T21:23:31Z scymtym: that is (as expected) identical to what declaring notinline make-instance produced
2016-08-11T21:24:25Z stassats: there are several that affected make-instance
2016-08-11T21:24:49Z stassats: 1) top-level are not optimized anymore 2) nil is not thrown out 3) gcing
2016-08-11T21:25:18Z scymtym: it's mostly 1) in this case
2016-08-11T21:36:25Z fe[nl]ix quit (Remote host closed the connection)
2016-08-11T21:36:25Z Blkt quit (Read error: Connection reset by peer)
2016-08-11T21:37:43Z Blkt joined #sbcl
2016-08-11T21:38:02Z fe[nl]ix joined #sbcl
2016-08-11T21:46:56Z stassats quit (Ping timeout: 258 seconds)
2016-08-11T21:53:22Z stassats joined #sbcl
2016-08-11T22:05:43Z chris2 quit (Ping timeout: 258 seconds)
2016-08-11T22:07:57Z chris2 joined #sbcl
2016-08-11T22:27:16Z Carisius quit (Read error: Connection reset by peer)
2016-08-11T22:27:30Z drmeister: Hi - does anyone know where these sbcl errors are coming from?
2016-08-11T22:27:48Z drmeister: https://www.irccloud.com/pastebin/PBwtU1ui/
2016-08-11T22:28:05Z drmeister: I cleared out the cache with:     rm -rf ~/.cache/common-lisp
2016-08-11T22:29:15Z drmeister: No matter what I do now,  sbcl goes nuts.
2016-08-11T22:30:13Z drmeister: Here's another one:
2016-08-11T22:30:15Z drmeister: https://www.irccloud.com/pastebin/D5mAbByQ/
2016-08-11T22:30:23Z drmeister: I guess I'll reinstall sbcl
2016-08-11T22:30:59Z stassats: are you corrupting the heap?
2016-08-11T22:31:34Z nyef: Weirdness in the .sbclrc, maybe?
2016-08-11T22:33:55Z fiddlerwoaroof joined #sbcl
2016-08-11T22:38:44Z drmeister: Maybe corrupting the heap - I don't think I have a .sbclrc
2016-08-11T22:38:47Z drmeister: Hi nyef.
2016-08-11T22:40:19Z nyef: Hello drmeister.
2016-08-11T22:42:18Z SamSkulls joined #sbcl
2016-08-11T22:42:22Z drmeister: I reinstalled sbcl and the problem went away.
2016-08-11T22:42:48Z stassats: a corrupt core file?
2016-08-11T22:57:18Z drmeister: Do the core files get put somewhere else?
2016-08-11T22:57:34Z drmeister: Doesn't matter - the problem went away.
2016-08-11T22:57:40Z stassats: as opposed to where?
2016-08-11T23:23:07Z angavrilov quit (Remote host closed the connection)
2016-08-11T23:25:01Z karswell joined #sbcl
2016-08-11T23:42:03Z DeadTrickster quit (Ping timeout: 276 seconds)
2016-08-11T23:50:39Z SamSkulls left #sbcl