2016-08-31T00:05:54Z slyrus quit (Ping timeout: 244 seconds)
2016-08-31T00:16:30Z oleo quit (Read error: Connection reset by peer)
2016-08-31T00:16:43Z oleo joined #sbcl
2016-08-31T00:30:55Z scymtym quit (Ping timeout: 255 seconds)
2016-08-31T01:18:18Z rpg joined #sbcl
2016-08-31T01:19:53Z sjl quit (Read error: Connection reset by peer)
2016-08-31T01:35:54Z DeadTrickster quit (Ping timeout: 265 seconds)
2016-08-31T01:45:57Z DeadTrickster joined #sbcl
2016-08-31T01:49:46Z stassats quit (Ping timeout: 252 seconds)
2016-08-31T01:54:14Z em1l_ joined #sbcl
2016-08-31T01:57:12Z em1l quit (Ping timeout: 240 seconds)
2016-08-31T02:10:56Z rpg quit (Quit: My Mac has gone to sleep. ZZZzzz…)
2016-08-31T03:08:17Z fisxoj quit (Quit: fisxoj)
2016-08-31T03:30:30Z DavidGu quit (Quit: DavidGu)
2016-08-31T03:41:09Z DavidGu joined #sbcl
2016-08-31T03:46:26Z rpg joined #sbcl
2016-08-31T03:55:39Z rpg quit (Quit: Textual IRC Client: www.textualapp.com)
2016-08-31T04:17:02Z edgar-rf_ joined #sbcl
2016-08-31T04:17:03Z edgar-rft quit (Ping timeout: 240 seconds)
2016-08-31T04:19:13Z edgar-rf_ is now known as edgar-rft
2016-08-31T04:21:32Z DeadTrickster quit (Ping timeout: 240 seconds)
2016-08-31T04:49:00Z oleo quit (Quit: Leaving)
2016-08-31T05:08:31Z muyinliu joined #sbcl
2016-08-31T05:12:27Z shka_ joined #sbcl
2016-08-31T06:25:19Z shka_ quit (Ping timeout: 252 seconds)
2016-08-31T06:30:51Z slyrus joined #sbcl
2016-08-31T07:05:26Z muyinliu quit (Quit: Textual IRC Client: www.textualapp.com)
2016-08-31T07:06:03Z muyinliu joined #sbcl
2016-08-31T07:06:08Z muyinliu: Can anyone tell me how to use SBCL's ldb?
2016-08-31T07:09:24Z scymtym joined #sbcl
2016-08-31T07:12:03Z DavidGu quit (Ping timeout: 240 seconds)
2016-08-31T07:19:56Z Xof: muyinliu: why do you want to use it?
2016-08-31T07:20:06Z Xof: (for what purpose?)
2016-08-31T07:20:57Z muyinliu: Xof: I wrote cl-graphicsmagick, it works fine in Mac OS X and CentOS 32bit systems, but crash in CentOS 64bit systems.
2016-08-31T07:21:26Z muyinliu: Xof: Pasted here: http://pasteall.org/77040
2016-08-31T07:26:19Z Bike: did you try initializing magick
2016-08-31T07:27:55Z Xof: ok.  It's unlikely that ldb will tell you very much about that error; it's a self-consistency check regarding signal handling that has got you into it
2016-08-31T07:28:08Z Xof: does graphicsmagick install any signal handlers?
2016-08-31T07:28:21Z Xof: (Unix signals, that is, not CL conditions)
2016-08-31T07:28:53Z muyinliu: Bike: You really should look at the file wand/magick_wand.c of GraphicsMagick, I pasted the part about NewMagickWand here: http://pasteall.org/77080/c. I'm sure GraphicsMagickWand initialize itself.
2016-08-31T07:29:56Z Bike: i'm just goin off the docs. i think it also said to call it before doing any threading, so if this is in slime that could be an issue?
2016-08-31T07:30:36Z muyinliu: Xof: ignal handlers? What do your mean?
2016-08-31T07:30:50Z muyinliu: Xof: Signal handlers? What do your mean?
2016-08-31T07:31:01Z Bike: like, posix signal handlers?
2016-08-31T07:32:05Z muyinliu: Bike: Sorry, I'm not familiar with signal handlers. But I think the code should NOT trigger memory fault...
2016-08-31T07:32:36Z Bike: https://en.wikipedia.org/wiki/Unix_signal
2016-08-31T07:34:44Z Xof: memory faults are not the only kind of signals
2016-08-31T07:34:58Z Xof: floating point operations can signal, for example
2016-08-31T07:35:00Z muyinliu: Bike: Thanks. I'm not sure whether you guys are talking about http://www.graphicsmagick.org/api/types.html#exceptiontype
2016-08-31T07:35:39Z Bike: it probably does not use signals for errors like that, no
2016-08-31T07:36:19Z muyinliu: Bike: sadly.
2016-08-31T07:40:32Z muyinliu: Bike: Xof: how to use SBCL's ldb?
2016-08-31T07:40:50Z Bike: if xof says it won't help i believe him
2016-08-31T07:42:10Z muyinliu: It seems that ldb can only print backtrace and the object of specific memory address
2016-08-31T07:43:04Z muyinliu: What is `pc` mean in the backtrace result?
2016-08-31T07:43:22Z muyinliu: And what is `fp` mean in the backtrace result ?
2016-08-31T07:46:51Z Xof: muyinliu: before you get all excited about ldb, please try to understand the error message before it: "blockables unblocked"
2016-08-31T07:47:09Z Xof: that means that some blockable unix signals, which SBCL expected to be blocked, are in fact unblocked
2016-08-31T07:47:22Z Xof: I think it's unlikely that messing around looking at the backtrace will tell you why this has happened
2016-08-31T07:48:08Z Bike: «InitializeMagick now registers signal handlers to ensure that resources are released before program exit. This helps avoid temporary file leaks due to the user using "CONTROL-C".»
2016-08-31T07:48:40Z muyinliu: Xof: Thanks.
2016-08-31T07:52:13Z muyinliu: Bike: I'm sure it's NOT about InitializeMagick...
2016-08-31T07:52:39Z Bike: no, this would be different from what i've been haranging you about
2016-08-31T07:52:56Z Bike: if initializemagick is called, it installs signal handlers, which messes with sbcl apparently
2016-08-31T07:54:07Z muyinliu: Bike: But why cl-graphicsmagick works fine on Mac OS X and CentOS 32bit? It makes no sense.
2016-08-31T07:54:27Z muyinliu: Bike: cl-graphicsmagick only crash on CentOS 64bit systems.
2016-08-31T07:54:34Z Bike: dunno
2016-08-31T07:54:46Z Bike: http://hg.code.sf.net/p/graphicsmagick/code/file/20f96db7e5a1/magick/magick.c#l1031 anyway it registers handlers for a bunch of stuff, maybe
2016-08-31T07:55:31Z Bike quit (Quit: asleep)
2016-08-31T07:59:52Z schjetne joined #sbcl
2016-08-31T08:01:50Z slyrus quit (Ping timeout: 258 seconds)
2016-08-31T08:05:49Z jsnell_: maybe the 64-bit machine has a different version of imagemagick installed?
2016-08-31T08:19:00Z DavidGu joined #sbcl
2016-08-31T08:23:15Z muyinliu: jsnell_: I'm sure libGraphicsMagickWand.so is OK. Because I test it with C program and it works fine. So the issue might comes from CFFI.
2016-08-31T08:23:23Z muyinliu quit (Quit: My Mac has gone to sleep. ZZZzzz…)
2016-08-31T08:26:20Z myrkraverk quit (Ping timeout: 265 seconds)
2016-08-31T08:27:37Z jsnell_: no, the issue is that imagemagick is enabling signals that sbcl expects to be in control of
2016-08-31T08:27:59Z jsnell_: you're refusing to believe this because it works for you on other machines
2016-08-31T08:28:32Z jsnell_: I am offering an alternative reason for why it's working on those other machines: they are running a different version of imagemagick that does not install those signal handlers
2016-08-31T08:29:30Z myrkraverk joined #sbcl
2016-08-31T08:40:44Z angular_mike quit (Ping timeout: 260 seconds)
2016-08-31T08:46:50Z angular_mike joined #sbcl
2016-08-31T08:59:03Z muyinliu joined #sbcl
2016-08-31T09:42:38Z prxq joined #sbcl
2016-08-31T09:49:03Z schjetne quit (Ping timeout: 264 seconds)
2016-08-31T09:57:02Z DeadTrickster joined #sbcl
2016-08-31T10:08:33Z schjetne joined #sbcl
2016-08-31T10:41:51Z schjetne quit (Ping timeout: 264 seconds)
2016-08-31T11:13:24Z DavidGu quit (Remote host closed the connection)
2016-08-31T11:22:15Z schjetne joined #sbcl
2016-08-31T11:22:16Z Posterdati quit (Read error: Connection reset by peer)
2016-08-31T11:24:09Z Posterdati joined #sbcl
2016-08-31T11:24:31Z whiteline quit (Ping timeout: 252 seconds)
2016-08-31T11:25:19Z muyinliu quit (Quit: Textual IRC Client: www.textualapp.com)
2016-08-31T11:37:59Z whiteline joined #sbcl
2016-08-31T11:42:48Z schjetne quit (Read error: Connection reset by peer)
2016-08-31T11:57:51Z sjl joined #sbcl
2016-08-31T12:13:01Z solene: hello, I am trying to compile sbcl on OpenBSD 6.0, which seems to be problematic now. I have been able to get a sbcl but it fails a test in sb-posix , here is the compile log, https://justpaste.it/xw1f
2016-08-31T12:13:09Z solene: I have Test SB-POSIX-TESTS::FCNTL.FLOCK.2 failed which fail
2016-08-31T12:31:31Z stassats joined #sbcl
2016-08-31T13:09:01Z salva quit (Ping timeout: 252 seconds)
2016-08-31T13:10:38Z rpg joined #sbcl
2016-08-31T13:12:10Z salva joined #sbcl
2016-08-31T13:29:54Z oleo joined #sbcl
2016-08-31T13:36:35Z _PosterdatiMobil joined #sbcl
2016-08-31T13:36:35Z PosterdatiMobile joined #sbcl
2016-08-31T13:38:55Z _PosterdatiMobil quit (Client Quit)
2016-08-31T13:39:56Z fisxoj joined #sbcl
2016-08-31T13:47:10Z PosterdatiMobile quit (Ping timeout: 250 seconds)
2016-08-31T13:52:10Z PosterdatiMobile joined #sbcl
2016-08-31T13:52:12Z _PosterdatiMobil joined #sbcl
2016-08-31T13:54:07Z _PosterdatiMobil quit (Client Quit)
2016-08-31T14:04:11Z cromachina quit (Read error: Connection reset by peer)
2016-08-31T14:06:34Z SamSkulls joined #sbcl
2016-08-31T14:09:42Z solene: after 15 days of work, I finally got sbcl working. but when I use the binary I need to add --core /path/to/sbcl.core
2016-08-31T14:10:00Z solene: is it possible to include inside sbcl binary the path to sbcl.core ?
2016-08-31T14:10:19Z stassats: yes, it's a build option
2016-08-31T14:10:39Z stassats: did it take you 15 days to type "--core /path/to/sbcl.core"?
2016-08-31T14:11:10Z solene: stassats: no, it took me 15 days to make it works on openbsd 6.0 due to some extra security features added which make sbcl not compliant with it
2016-08-31T14:11:40Z solene: clisp is slow to bootstrap sbcl :(
2016-08-31T14:14:22Z scymtym: stassats: current progress: http://paste.lisp.org/display/324705 . it's a draft, don't look at the details :)
2016-08-31T14:14:53Z scymtym: do you agree with the general approach and would you help with the backend changes?
2016-08-31T14:17:45Z stassats: a solution to assem.S would be exporting the encoded scoffsets from lisp
2016-08-31T14:19:17Z scymtym: as a precomputed lookup table instead of the BYTE-like specification i used in patch [3/3]?
2016-08-31T14:20:09Z stassats: well, no, just the arguments it needs for UNDEFINED_FUN_ERROR
2016-08-31T14:20:45Z scymtym: oh, i didn't flesh this out yet, but struct sc_offset_byte sc_offset_sc_number_bytes[] and friends are supposed to be dumped in genesis.lisp
2016-08-31T14:20:46Z stassats: or maybe some cpp wizardry
2016-08-31T14:21:04Z scymtym: there was a FIXME comment for that
2016-08-31T14:21:09Z scymtym: *is
2016-08-31T14:21:59Z solene: I have a weird problem. My system sbcl works (if I start it as root, it works as expected, with the core file loaded, just by typing sbcl), and with my regular user, whcih with I compiled sbcl from hand in a directory, sbcl tries to load /home/solene/Downloads/sbcl/lib/sbcl/sbcl.core which is not the right sbcl.core file, and I just type "sbcl". Why is it trying to load this ?
2016-08-31T14:22:41Z solene: found => SBCL_HOME=/home/solene/Downloads/sbcl/lib/sbcl sorry
2016-08-31T14:25:14Z stassats: !define-storage-classes should be shared by all the backends
2016-08-31T14:25:27Z SamSkulls quit (Ping timeout: 265 seconds)
2016-08-31T14:26:01Z scymtym: i can try to do that as [0/3]
2016-08-31T14:29:29Z scymtym: the version in hppa/vm.lisp even has a comment ";;; FIXME-lav: move this into arch-generic-helpers.lisp and rip out from arches"
2016-08-31T14:35:47Z stassats: nyef changes to move all the tramps into read-only space would help with encoding this stuff
2016-08-31T14:36:43Z scymtym: stassats: thanks for your thoughts. i'm out of time for now. will try to continue in the evening (and read the logs)
2016-08-31T14:42:07Z stassats: ok, i think read-only-tramps should be integrated first
2016-08-31T14:43:11Z scymtym quit (Ping timeout: 258 seconds)
2016-08-31T14:50:39Z nyef: ... Oh, right. We need to move the tramps to read-only space on HPPA anyway.
2016-08-31T14:50:54Z nyef: That's on my list to do somewhere.
2016-08-31T14:51:17Z stassats: i can cover x86oids and armoids
2016-08-31T14:51:20Z stassats: and mips, kinda
2016-08-31T14:51:29Z stassats: (doesn't survive cold init)
2016-08-31T14:53:27Z nyef: MIPS doesn't survive cold init? What?
2016-08-31T14:53:56Z nyef: I can cover everything but PPC and ARMoids, and possibly even PPC and ARM32.
2016-08-31T14:54:13Z stassats: my thoughts exactly
2016-08-31T14:54:33Z nyef: I thought that the problem with MIPS on the ML was that it was a non-FPU box?
2016-08-31T14:55:12Z stassats: that wasn't my problem
2016-08-31T14:55:20Z nyef: Ah, okay.
2016-08-31T14:55:31Z nyef: Sounds like I need to do a full round on my build farm.
2016-08-31T14:56:36Z stassats: actually, it doesn't survive make-target-2
2016-08-31T14:57:12Z stassats: debugger invoked on a SB-INT:BUG:
2016-08-31T14:57:12Z stassats:   #S(SB-INT:BUG
2016-08-31T14:57:13Z stassats:      :ACTUAL-INITARGS (FORMAT-CONTROL ~@<failed AVER: ~2I~_~S~:>
2016-08-31T14:57:15Z stassats:                        FORMAT-ARGUMENTS ((NOT (BLOCK-DELETE-P BLOCK))))
2016-08-31T14:57:18Z stassats:      :ASSIGNED-SLOTS NIL)
2016-08-31T14:57:23Z nyef: That... doesn't look good.
2016-08-31T14:58:22Z nyef wakes up his MIPS box.
2016-08-31T14:58:37Z stassats: that was be
2016-08-31T14:58:57Z nyef: Not sure I have a usable mipsle box, tbh.
2016-08-31T14:59:39Z stassats: le actually works
2016-08-31T15:01:36Z stassats: building the latest git
2016-08-31T15:01:54Z nyef: Doing a git pull now.
2016-08-31T15:02:35Z nyef: ... I don't think that I deployed the kernel with the pagetable fix to this machine, but that only really affects the guard pages.
2016-08-31T15:03:56Z nyef: Building now.
2016-08-31T15:04:02Z stassats: Linux erpro8-fsf1 3.14.10-er8mod-00013-ge0fe977 #1 SMP PREEMPT Wed Jan 14 12:33:22 PST 2015 mips64 GNU/Linux
2016-08-31T15:04:04Z stassats: on the broken be
2016-08-31T15:04:15Z stassats: Linux erpro8-fsf3 4.1.4 #1 SMP PREEMPT Mon Aug 3 14:22:54 PDT 2015 mips64 GNU/Linux
2016-08-31T15:04:18Z stassats: on the working le
2016-08-31T15:06:35Z nyef: [    0.000000] Linux version 4.2.0-00085-g4e9e5b3 (nyef@kana) (gcc version 4.7.4 (Buildroot 2014.11) ) #622 Mon Oct 26 12:24:44 EDT 2015
2016-08-31T15:06:35Z nyef:  
2016-08-31T15:07:19Z nyef: Trying to get a "proper" uname -a now.
2016-08-31T15:07:44Z stassats: since le works, i can handle changes to mips as well
2016-08-31T15:07:53Z nyef: I think that the SBCL build is taking up too much of the CPU, though, I can't even connect a second time. /-:
2016-08-31T15:07:55Z stassats: leaving alphppa uncovered
2016-08-31T15:08:30Z nyef: alpha is, IIRC, still brokenish, due to runtime DX SAPs and the half-wi(d)t(h) heap.
2016-08-31T15:11:20Z nyef: hppa is also still brokenish, I still hadn't finished bringing it up to scratch when I left off for various reasons.
2016-08-31T15:12:07Z nyef: Oh! The changes to move the trampolines to read-only space were feature-conditioned anyway, weren't they?
2016-08-31T15:12:20Z stassats: yes
2016-08-31T15:12:49Z stassats: but i wanted complete coverage for the sc-offset changes
2016-08-31T15:13:31Z nyef: Sure, that I understand.
2016-08-31T15:13:45Z nyef: Umm... Are we still in code-freeze?
2016-08-31T15:13:58Z nyef: Didn't even look at the logs before building.
2016-08-31T15:14:10Z stassats: we aren't
2016-08-31T15:14:16Z nyef: Okay then.
2016-08-31T15:22:41Z nyef: Still in host-2.
2016-08-31T15:23:58Z Bike joined #sbcl
2016-08-31T15:38:10Z stassats: #<TYPE-ERROR (:DATUM NIL :EXPECTED-TYPE (MEMBER T NIL)) {31EA9129}>
2016-08-31T15:38:14Z stassats: so, that's on be
2016-08-31T15:38:25Z stassats: COMPILE-FILE "SYS:SRC;PCL;MACROS.LISP"
2016-08-31T15:38:39Z nyef: Still in host-2 for me.
2016-08-31T15:39:06Z stassats: which is strange, that's the top frame
2016-08-31T15:39:49Z stassats: there are two object-not-boolean-error in compile-file
2016-08-31T15:41:19Z stassats: probably for warnings-p
2016-08-31T15:41:19Z stassats:             failure-p
2016-08-31T15:41:53Z attila_lendvai joined #sbcl
2016-08-31T15:41:53Z attila_lendvai quit (Changing host)
2016-08-31T15:41:53Z attila_lendvai joined #sbcl
2016-08-31T15:43:54Z nyef: If necessary, I can try to bisect to find the build failure.
2016-08-31T15:43:59Z edgar-rft quit (Quit: edgar-rft)
2016-08-31T15:44:11Z nyef: Presuming, of course, that I manage to GET the build failure.
2016-08-31T15:44:21Z stassats: i have i feeling it only fails for me
2016-08-31T15:44:58Z nyef: We'll find out if it fails for me sooner or later.
2016-08-31T15:45:18Z nyef: cross-compiling compiler/ctype now.
2016-08-31T15:45:53Z nyef: Maybe 2/3 through host-2?
2016-08-31T15:47:43Z nyef: Ah! I can log in through the console, even if ssh is timing out for some reason.
2016-08-31T15:47:54Z nyef: Linux cf00 4.2.0-00085-g4e9e5b3 #622 Mon Oct 26 12:24:44 EDT 2015 mips64 R16000 V4.1 FPU V0.0 SGI Origin 300/350 GNU/Linux
2016-08-31T15:49:27Z nyef: Okay, there's no way in hell this has the FPU fixes for FR1 mode or the page table fix.
2016-08-31T15:50:18Z nyef: Also doesn't have SMP, but I don't think that anybody has SMP on Linux on this hardware.
2016-08-31T15:51:18Z stassats: why would a type check fail on big-endian
2016-08-31T15:51:48Z nyef: Someone adjusting the type-check code for something, and missing an endian compensation.
2016-08-31T15:51:51Z christoph_debian: so mipsel is working these days?
2016-08-31T15:52:00Z nyef: If it's generic code, it would fail on PPC as well, surely?
2016-08-31T15:52:04Z christoph_debian: and in somewhat maintained state
2016-08-31T15:52:10Z nyef: Also some ARM boards.
2016-08-31T15:52:22Z christoph_debian: last time I checked (quite a while ago) it never managed to build contrib afair
2016-08-31T15:52:24Z stassats: nyef: nobody runs on those, though
2016-08-31T15:52:43Z nyef: Which, ARM BE, or PPC?
2016-08-31T15:52:47Z stassats: arm be
2016-08-31T15:54:03Z stassats: an sb-posix test fails on mips, probably a broken test
2016-08-31T15:54:17Z fisxoj quit (Quit: fisxoj)
2016-08-31T15:54:56Z nyef: The main failure that I remember in the test suite is the stack exhaustion test causing the entire process to disappear, but that turned out to be a kernel bug that only kicked in on some hardware.
2016-08-31T15:55:08Z nyef: Was there for YEARS before we tracked it down. (-:
2016-08-31T15:55:09Z stassats: considering that it succeeds when i run it
2016-08-31T15:56:06Z nyef: The fix is in the latest kernel or two, and it only triggers under certain cache coherency schemes.
2016-08-31T15:56:30Z nyef: (Might not have been cache coherency schemes, but something like that. Not a common vulnerability, at any rate.)
2016-08-31T15:59:39Z stassats: ok, it fails separately too
2016-08-31T15:59:46Z stassats: was running the wrong one
2016-08-31T16:02:50Z christoph_debian: jep that one.
2016-08-31T16:03:33Z stassats: ok, it's a bogus test alright
2016-08-31T16:03:53Z nyef: Hrm... I forget, is my VPN host BE or LE?
2016-08-31T16:03:57Z stassats: or a bogus kernel
2016-08-31T16:04:01Z stassats: it checks the flock's pid
2016-08-31T16:04:12Z stassats: this is root     28287  0.0  0.0      0     0 ?        S<   Aug09  11:06 [kworker/0:1H] the pid it returns
2016-08-31T16:06:13Z nyef: Oh, hell. My VPN host isn't responding.
2016-08-31T16:07:35Z stassats: so, just slap #-mips on it?
2016-08-31T16:12:15Z nyef: VPN heatsink is hot. Let's eliminate the airflow around it by placing it smack against the case for the Alpha... And let it cool down a bit before trying again.
2016-08-31T16:15:09Z stassats: nyef: inform me whether sb-posix fcntl.flock.2 fails for you too
2016-08-31T16:15:25Z nyef: If my build gets that far, sure.
2016-08-31T16:16:05Z stassats: ok, i'm done with mips for the day
2016-08-31T16:21:19Z nyef: "note: doing signed-word to integer coercion (cost 18), for: the first argument of FIXNUMP"?
2016-08-31T16:22:47Z nyef: Ah! Genesis, at last.
2016-08-31T16:34:09Z nyef: Building contribs.
2016-08-31T16:35:25Z stassats: yeah...
2016-08-31T16:36:53Z nyef: Looks like your build failure IS machine-specific.
2016-08-31T16:44:50Z nyef: Preliminary results are no test failures on sb-posix, but the serial console has a messed up terminal, so I'm waiting for the build to finish to check for that specific test.
2016-08-31T16:48:29Z nyef: Build complete, elapsed time... about an hour and three quarters.
2016-08-31T16:48:35Z nyef: That seems long, somehow.
2016-08-31T16:48:50Z flip214: and sounds like an imperial measurement ;)
2016-08-31T16:49:35Z nyef: stassats: FCNTL.FLOCK.2 passed for me.
2016-08-31T16:51:01Z stassats: ok
2016-08-31T16:51:10Z sjl quit (Ping timeout: 244 seconds)
2016-08-31T16:51:12Z stassats: well, this one is clearly a broken kernel
2016-08-31T16:51:16Z stassats: but not enough for #-mips
2016-08-31T16:53:57Z nyef: Hrm. I bet the VPN box works better if I use the right IP for it. /-:
2016-08-31T16:55:21Z nyef: Preparing to do a build on my VPN box as well, for completenes sake.
2016-08-31T17:06:55Z nyef: So, how about those read-only tramps?
2016-08-31T17:12:04Z scymtym joined #sbcl
2016-08-31T17:32:39Z gingerale joined #sbcl
2016-08-31T18:08:18Z p_l quit (Ping timeout: 244 seconds)
2016-08-31T18:10:32Z sjl joined #sbcl
2016-08-31T18:11:52Z stassats: nyef: yes, read-only tramps
2016-08-31T18:12:17Z shka_ joined #sbcl
2016-08-31T18:12:33Z p_l joined #sbcl
2016-08-31T18:22:26Z stassats: trying to merge first
2016-08-31T18:31:05Z edgar-rft joined #sbcl
2016-08-31T18:49:54Z stassats: merged read-only tramps https://github.com/stassats/sbcl
2016-08-31T18:51:39Z stassats: into master, but ok
2016-08-31T18:52:39Z attila_lendvai quit (Ping timeout: 260 seconds)
2016-08-31T19:00:36Z stassats: COLD-INIT... (PACKAGE = "COMMON-LISP")
2016-08-31T19:00:36Z stassats: Heap exhausted during allocation: 32 bytes available, 48 requested.
2016-08-31T19:00:37Z stassats: hello
2016-08-31T19:04:51Z nyef: The required logic may have changed slightly?
2016-08-31T19:05:06Z stassats: i'm adjusting x86-64, so, didn't adjust properly
2016-08-31T19:05:18Z stassats: now it lacks :assembly-routine fixup
2016-08-31T19:05:43Z stassats: weird
2016-08-31T19:08:16Z stassats: looks like a sanity check
2016-08-31T19:10:23Z mason quit (Quit: leaving)
2016-08-31T19:10:41Z stassats: failed AVER: (AND SB-IMPL::TABLE-ADDRESS SB-IMPL::REAL-ADDRESS)
2016-08-31T19:10:53Z stassats: (SB-SYS:UPDATE-LINKAGE-TABLE)
2016-08-31T19:10:53Z stassats: ok
2016-08-31T19:12:08Z mason joined #sbcl
2016-08-31T19:23:50Z stassats: ok, contribbing
2016-08-31T19:24:25Z stassats: (a) => debugger invoked on a UNDEFINED-FUNCTION in thread => 0: (SB-VM::UNDEFINED-TRAMP)
2016-08-31T19:26:04Z stassats: sb-mpfr failed to build? huh
2016-08-31T19:40:44Z trinque: https://en.wikipedia.org/wiki/SourceForge#Project_hijackings_and_bundled_malware << I'm curious; why does sbcl continue to use sourceforge as primary repository?
2016-08-31T19:41:07Z stassats: because that's just some FUD
2016-08-31T19:41:25Z trinque: is it? I'm not aware
2016-08-31T19:41:45Z stassats: just because the internet hivemind decided that sourceforge is bad doesn't mean it's bad
2016-08-31T19:42:13Z trinque: did those events not happen?
2016-08-31T19:44:12Z trinque: relatedly, where can I find a gpg sig for a particular source release?
2016-08-31T19:44:47Z stassats: right beside it?
2016-08-31T19:45:26Z trinque: ah, I loverlooked it. thanks
2016-08-31T19:45:34Z trinque: *overlooked, hah
2016-08-31T19:47:36Z trinque: at any rate, since you ship binaries (and I assume the signature is for the sourceball) I'd think sf meddling with installers might be concerning.
2016-08-31T19:48:08Z trinque: or at least might want to provide sigs for those too
2016-08-31T19:48:42Z stassats: for all you know, i might be embedding malware on my own
2016-08-31T19:48:55Z trinque: thus why even use crypto, is your point?
2016-08-31T19:49:05Z stassats: it's who you trust
2016-08-31T19:49:17Z stassats: i'm not any more trustworthy than sourceforge
2016-08-31T19:49:40Z stassats: the x86-64 binary is signed
2016-08-31T19:49:40Z stassats: fwiw
2016-08-31T19:49:43Z stassats: linux
2016-08-31T19:51:43Z trinque: this is why establishing a web of trust is useful, which cannot be done if relationships are ambiguous.
2016-08-31T19:52:00Z stassats: and if you don't have Xof's key in your keyring, those signatures aren't worth much
2016-08-31T19:52:29Z Xof: the copy of my key that was on sourceforge had an error in it ever since it was uploaded
2016-08-31T19:52:31Z stassats: frankly, i don't believe in the web of trust, you can only know who uploaded
2016-08-31T19:52:35Z stassats: not their intentions
2016-08-31T19:52:40Z Xof: it took the Internet years to notice.  (I didn't do it deliberately)
2016-08-31T19:52:56Z stassats: i may be offered a hefty sum to sign with my verified signature something malicious
2016-08-31T19:52:57Z trinque: stassats: you at least know who to blame
2016-08-31T19:53:03Z trinque: this is worth quite a bit
2016-08-31T19:53:08Z stassats: doesn't matter, got money!
2016-08-31T19:53:25Z trinque: and when this happens, I can sign a statement saying it did
2016-08-31T19:53:28Z trinque: thus reputation builds
2016-08-31T19:53:48Z trinque: cannot happen if items are not signed; I see that they are, so I don't perceive we disagree on at least the utility of that
2016-08-31T19:54:11Z stassats: well, the binaries i upload are not signed
2016-08-31T19:55:02Z stassats: for all i know i may have a windows virus that infects stuff (i doubt that, it's a VM exclusively for building SBCL, but still)
2016-08-31T19:55:20Z trinque: this argument can be applied to not bathing
2016-08-31T19:55:30Z stassats: it can't
2016-08-31T19:56:40Z trinque: the form of it is "why do X when X does not obviate all possible failures"
2016-08-31T19:57:06Z stassats: well, it's my argument and i do not apply it to bathing
2016-08-31T19:57:14Z trinque: if the item is signed and my box is infected, I may come back to you and prove to you that it is the binary you gave me that did this
2016-08-31T19:58:11Z stassats: well, cool, but i'm not going start signing
2016-08-31T19:58:24Z stassats: build from sources
2016-08-31T19:59:33Z stassats: ok, successfully transplaned undefined, closure and undefined-alien into read-only on x86-64
2016-08-31T19:59:58Z stassats: arm64 next, and maybe sb-dynamic-core will start working there, that'll be grand
2016-08-31T20:01:53Z stassats: but first, funcallable-instance-tramp for x86-64
2016-08-31T20:02:43Z trinque: Xof: is this also your perspective on gpg signatures?
2016-08-31T20:02:51Z trinque: it's surprising.
2016-08-31T20:02:59Z stassats: well, Xof does sign
2016-08-31T20:03:03Z Xof: they have their uses
2016-08-31T20:03:36Z Xof: I would hope that e.g. debian sbcl maintainers check my signature, if they use my source tarball, but that largely works because I have met them
2016-08-31T20:03:39Z stassats: if signing stuff required zero effort i would do it
2016-08-31T20:04:39Z trinque: Xof: that is the only way it works
2016-08-31T20:04:46Z Xof: does the fact that I get gpg-signed updates over apt from debian reassure me?  Not particularly, because around 1000 people can make changes to the packages that I get
2016-08-31T20:04:54Z trinque: certainly not
2016-08-31T20:05:54Z trinque: but yes, if I can find someone who has met you (which isn't unreasonable) in my WoT, I can be reasonably certain I'm getting the sbcl with which your reputation is associated
2016-08-31T20:05:57Z rpg quit (Ping timeout: 244 seconds)
2016-08-31T20:06:09Z trinque: this is as valuable to me as to debian
2016-08-31T20:06:42Z stassats: well, you have signed sources, signed x86-64-linux binary
2016-08-31T20:06:49Z stassats: you can build and cross build all other platforms
2016-08-31T20:26:02Z stassats: succeess with funcallable-instance-tramp
2016-08-31T20:27:48Z akkad left #sbcl
2016-08-31T20:30:04Z gingerale quit (Remote host closed the connection)
2016-08-31T21:01:03Z sjl quit (Ping timeout: 240 seconds)
2016-08-31T21:07:42Z christoph_debian: Xof: fwiw I check the signature on the release tag
2016-08-31T21:08:28Z christoph_debian: against a limited keyring
2016-08-31T21:11:56Z christoph_debian: and while I'm to one doing the debian packages for years now I believe we never met in person? though I would certainly enjoy
2016-08-31T21:12:24Z jackdaniel: ^_^
2016-08-31T21:12:25Z Xof: that might be the case
2016-08-31T21:12:55Z scymtym: stassats: assuming that error breaks never receive more than four arguments, i could squeeze the number of arguments into the first sc-offset in common cases instead of using a dedicating a byte to it
2016-08-31T21:14:35Z Xof: of the historical sbcl debian maintainers, I have met several
2016-08-31T21:14:48Z stassats: tagged tramps are more difficult
2016-08-31T21:14:49Z christoph_debian: (-:
2016-08-31T21:15:09Z Xof: (including the guy all the way at the bottom of debian/changelog)
2016-08-31T21:16:31Z stassats: hm, i can just trick my way into it for now
2016-08-31T21:17:07Z stassats: instead of having a label-3, i can load the label into the register and subtract 3, and deal with optimizing it later
2016-08-31T21:17:44Z shka_ quit (Ping timeout: 265 seconds)
2016-08-31T21:18:25Z stassats: oh wait, it won't be tagged, blimey
2016-08-31T21:18:44Z stassats: but it's in read-only space, so who cares
2016-08-31T21:19:16Z stassats: or maybe it's important for interrupts
2016-08-31T21:19:17Z stassats: sigh
2016-08-31T21:23:09Z Xof: running tests takes a long time on rpi
2016-08-31T21:27:16Z Carisius joined #sbcl
2016-08-31T21:27:28Z Carisius quit (Remote host closed the connection)
2016-08-31T21:28:31Z nyef: stassats: Part of the read-only-tramps work is enabling assembly-routines to have symbols with an offset against a label.
2016-08-31T21:29:07Z stassats: i just added an optional offsets for the ADR instruction on ARM
2016-08-31T21:30:18Z nyef: ... In fact, this functionality was required for the PPC tramps.
2016-08-31T21:30:42Z prxq: Xof: the recommendation for that kind of task (testing code for embedded devices) is using an emulator.
2016-08-31T21:31:01Z stassats: which is even slower?
2016-08-31T21:31:08Z stassats: and doesn't test properly?
2016-08-31T21:31:08Z prxq: normally not.
2016-08-31T21:31:17Z prxq: normally not.
2016-08-31T21:31:18Z nyef: As an emulator author, I do not trust emulators.
2016-08-31T21:31:53Z stassats: i'd wager rpi is faster than any emulator
2016-08-31T21:34:46Z Xof: on the all-important "build sbcl" benchmark, it is around 25 times slower than my 5-year-old laptop
2016-08-31T21:36:06Z Xof: wait, does that make sense?
2016-08-31T21:36:31Z stassats: sounds right
2016-08-31T21:43:08Z ASau joined #sbcl
2016-08-31T21:47:18Z rpg joined #sbcl
2016-08-31T21:50:06Z angavrilov_ quit (Remote host closed the connection)
2016-08-31T21:55:59Z stassats: arm64-readonly-tramps fails somewhere in make-target-2, probably funcallable-instance-tramp
2016-08-31T22:02:18Z stassats: pc at 000175, weird
2016-08-31T22:02:25Z stassats: the funcallable-instance-tramp is at 148
2016-08-31T22:03:24Z ASau quit (Remote host closed the connection)
2016-08-31T22:03:53Z nyef: Is there a tag bias being applied somewhere?
2016-08-31T22:03:57Z ASau joined #sbcl
2016-08-31T22:04:13Z stassats: well, apparently closure-tramp works alright
2016-08-31T22:04:26Z nyef: It's a different style of tramp.
2016-08-31T22:04:36Z nyef: ... isn't it?
2016-08-31T22:04:53Z stassats: i remember there being something different about it
2016-08-31T22:05:14Z stassats: ok, yeah
2016-08-31T22:12:52Z stassats: ok, now to figure how to translate that difference
2016-08-31T22:14:13Z ASau quit (Remote host closed the connection)
2016-08-31T22:14:34Z nyef: Is it any different than the difference on x86?
2016-08-31T22:15:25Z stassats: x86 do not tag tramps
2016-08-31T22:15:40Z nyef: Does x86-64?
2016-08-31T22:15:40Z ASau joined #sbcl
2016-08-31T22:15:47Z stassats: neither
2016-08-31T22:15:55Z nyef: Hrm.
2016-08-31T22:17:04Z nyef: Looks like the difference is that there's no header on the funcallable instance tramp?
2016-08-31T22:17:51Z stassats: well, there is in arm64-assem.S
2016-08-31T22:18:10Z nyef: ARM is different.
2016-08-31T22:18:15Z nyef: Ah!
2016-08-31T22:18:20Z nyef: Right, you're on ARM.
2016-08-31T22:18:28Z stassats: ok, funcallable-instance-tramp is the one that's tagged
2016-08-31T22:18:39Z stassats: all other tramps are not tagged
2016-08-31T22:19:12Z nyef: Right, the PPC backend is the same, so it should have a usable precedent.
2016-08-31T22:19:53Z stassats: well, except that ppc read-only-tramps do not have funcallable-instance-tramp
2016-08-31T22:20:13Z nyef: ... Er?
2016-08-31T22:20:24Z nyef: Oh, right. I only made it happen through genesis, didn't I?
2016-08-31T22:21:26Z nyef: Looks like you'll need the :ALIGNMENT and :EXPORT options to define-assembly-routine.
2016-08-31T22:21:28Z ASau quit (Remote host closed the connection)
2016-08-31T22:21:43Z stassats: i think i've got it figured out
2016-08-31T22:21:45Z stassats: building
2016-08-31T22:21:46Z nyef: Err... :ALIGN.
2016-08-31T22:22:04Z ASau joined #sbcl
2016-08-31T22:22:45Z stassats: i did figure something out, since i'm getting a different fault now
2016-08-31T22:22:50Z nyef: Oh, heh! The PPC undefined tramp defines and uses a fixup for its tagged pointer, because it's a recursive structure.
2016-08-31T22:22:59Z nyef: The simple-fun-self thing.
2016-08-31T22:23:38Z stassats: ok, the tramp is called alright
2016-08-31T22:23:45Z stassats: it's just wrong
2016-08-31T22:24:18Z nyef: I should be able to cover SPARC, HPPA, Alpha (probably), and MIPS BE.
2016-08-31T22:25:00Z stassats: i can test mips le, or do mips, since it shouldn't be any different
2016-08-31T22:25:12Z stassats: and complete ppc, and x86 and add arm
2016-08-31T22:25:51Z stassats: huh, lexenv-tn has some bogus value
2016-08-31T22:26:18Z nyef: register assignment difference between lisp and runtime?
2016-08-31T22:26:33Z stassats: hardcoded
2016-08-31T22:26:55Z nyef: Broken closure tramp?
2016-08-31T22:27:14Z stassats: and i'm looking in gdb
2016-08-31T22:27:24Z stassats: so, the offsets are right
2016-08-31T22:27:27Z stassats: just something bogus
2016-08-31T22:29:14Z stassats: oh, wrong lowtag
2016-08-31T22:29:17Z stassats: i think
2016-08-31T22:29:30Z stassats: wouldn't explain the error, but it's wrong anyway
2016-08-31T22:31:12Z stassats: well, i'm past braid.lisp
2016-08-31T22:31:26Z prxq quit (Remote host closed the connection)
2016-08-31T22:31:38Z stassats: and it's warm and clossy
2016-08-31T22:35:40Z stassats: strange, i don't see (undefined-tramp) in the backtrace
2016-08-31T22:38:12Z stassats: looks like non-x86oids have better undefined-function detection
2016-08-31T22:40:21Z stassats: pushed everything to https://github.com/stassats/sbcl
2016-08-31T22:40:30Z stassats: and that's it for read-only-tramps for today
2016-08-31T22:41:38Z stassats dreams of the day when there is no arch-assem.S
2016-08-31T22:54:21Z nyef: We... can make that happen, mostly.
2016-08-31T22:54:37Z stassats: i think that's what stopped me from porting to AIX
2016-08-31T23:21:18Z nyef: I can make mips-assem.S go away.
2016-08-31T23:24:31Z nyef: And most of ppc-assem.S.
2016-08-31T23:48:25Z ASau quit (Remote host closed the connection)
2016-08-31T23:50:56Z ASau joined #sbcl
2016-08-31T23:51:36Z sjl joined #sbcl