2014-11-14T00:31:43Z joshe joined #sbcl
2014-11-14T00:38:08Z psilord joined #sbcl
2014-11-14T00:46:18Z oleo__ joined #sbcl
2014-11-14T00:47:43Z oleo is now known as Guest22182
2014-11-14T00:48:17Z Guest22182 quit (Ping timeout: 240 seconds)
2014-11-14T01:05:47Z krzysz00 joined #sbcl
2014-11-14T01:17:51Z ehaliewicz quit (Remote host closed the connection)
2014-11-14T01:18:28Z ehaliewicz joined #sbcl
2014-11-14T01:26:05Z ehaliewicz quit (Remote host closed the connection)
2014-11-14T01:26:21Z ehaliewicz joined #sbcl
2014-11-14T01:55:49Z stassats: sprof seems to be good at doing GC invariant lost, gencgc.c:860
2014-11-14T01:59:21Z nyef: Oof. Is that the one that the region is in a reset state, or is that the mutex lock?
2014-11-14T02:01:29Z stassats quit (Ping timeout: 245 seconds)
2014-11-14T02:14:10Z nyef: minion: memo for stassats: You might find http://repo.or.cz/w/sbcl/nyef.git/shortlog/refs/heads/read-only-tramps to be interesting. Not ready for "prime time" yet, as it were, but getting there.
2014-11-14T02:14:11Z minion: Remembered. I'll tell stassats when he/she/it next speaks.
2014-11-14T02:17:55Z kanru` joined #sbcl
2014-11-14T02:18:05Z krzysz00: stassats: What sort of questions does targen-unicode raise?
2014-11-14T02:18:44Z krzysz00: Also, I seem to remember putting the normalization switch in sb-ext
2014-11-14T02:20:46Z krzysz00: Krystof: I figured we should do NFKC because CL is case-weird (but usually case-insensitive)
2014-11-14T02:36:11Z echo-area joined #sbcl
2014-11-14T02:50:02Z stassats joined #sbcl
2014-11-14T02:50:04Z stassats quit (Changing host)
2014-11-14T02:50:04Z stassats joined #sbcl
2014-11-14T02:50:07Z stassats quit (Client Quit)
2014-11-14T02:56:27Z ehaliewicz quit (Ping timeout: 272 seconds)
2014-11-14T03:04:49Z Bicyclidine quit (Ping timeout: 245 seconds)
2014-11-14T03:05:16Z joshe: hm, I appear to have a working pa-risc machine
2014-11-14T03:07:58Z joshe: the disk containing my 6-year-old openbsd/hppa port attempt seems to have died, which is not exactly a great loss
2014-11-14T03:20:38Z nyef: joshe: Reeealy...? Oddly enough, I, too, should have a working pa-risc machine, mine with a semi-recent Debian install.
2014-11-14T03:21:18Z jlarocco quit (Quit: This computer has gone to sleep)
2014-11-14T03:21:29Z joshe: my power supply or motherboard or something died in 2008 and I finally got around to buying a new one
2014-11-14T03:21:55Z joshe: does debian still support pa-risc?
2014-11-14T03:22:13Z nyef: And, speaking of, do you know what's involved in doing a CPU swap on one of these? I have a not-working machine with two higher-end CPUs than the one in my working machine, but they're both the same basic model machine.
2014-11-14T03:22:27Z joshe: that'll make it easier to bring up an openbsd port, when I replace the failed disk
2014-11-14T03:23:08Z joshe: I have no idea about a cpu swap, sorry
2014-11-14T03:23:38Z nyef: There's an "unofficial" port that still apparently gets some love.
2014-11-14T03:24:38Z jlarocco joined #sbcl
2014-11-14T03:24:56Z nyef: I'm also hoping to figure out what to do about getting a drive mounted in an Alpha DS20L. For some reason I keep ending up with alpha machines without the drive mounting hardware. /-:
2014-11-14T03:25:10Z joshe: heh
2014-11-14T03:25:47Z nyef: Worst-case scenario, I have it use a network block device of some sort and net-boot the sucker from my XServe G5.
2014-11-14T03:25:58Z joshe: I bought a nice fast 1u alpha on ebay recently, but apparently in the year between when they'd received it and when they found a sucker to buy it, the hardware had failed
2014-11-14T03:26:18Z nyef: Ouch.
2014-11-14T03:26:32Z nyef: Which model, the DS10L or the DS20L?
2014-11-14T03:26:38Z joshe: luckily they tested it again before shipping it
2014-11-14T03:26:41Z joshe: the 20
2014-11-14T03:26:56Z joshe: api cs20 or something like that
2014-11-14T03:26:58Z nyef: Mmm. Two CPUs, tolerably fast, and so on.
2014-11-14T03:27:38Z joshe: there's still a couple listed on ebay, but for more than I'm willing to spend
2014-11-14T03:27:52Z nyef: Yeah, that's the version I got as well. It was rebadged as the DS20L when DE-paq-HP also started selling it.
2014-11-14T03:28:40Z nyef: I also have a DS10L, but it's a single CPU and for some reason is also missing the drive tray.
2014-11-14T03:29:00Z joshe: sometimes I think I might have a problem
2014-11-14T03:29:08Z joshe: my ebay home screen is full of vaxen
2014-11-14T03:29:44Z nyef: I would have been tempted by a VAX at one point, but I really, REALLY don't feel like trying to port SBCL to a new CPU architecture any time soon. (-:
2014-11-14T03:30:59Z joshe: freebsd might be dropping itanium, but openbsd vax support is still improving ;)
2014-11-14T03:31:10Z joshe: it recently moved to ELF and gcc3
2014-11-14T03:31:56Z nyef: Another item on my project list that hasn't gone anywhere recently is getting Linux installed on my O350. And, yes, I know that openbsd should run on it.
2014-11-14T03:32:23Z nyef: Hrm.
2014-11-14T03:32:42Z joshe: I seem to recall some console problems with that machine
2014-11-14T03:32:44Z nyef: Do you have a decent openbsd mips box?
2014-11-14T03:32:49Z joshe: but I've never owned sgi hardware
2014-11-14T03:33:23Z joshe: I have a couple cheap octeon boxes that are supposed to netboot nowadays, but I haven't tried
2014-11-14T03:33:47Z nyef: I'm not too fussed about console, TBH. I've been collecting 1U and 2U rackmount stuff and figuring to just throw them on the localnet and access them over ssh.
2014-11-14T03:33:52Z joshe: however sbcl would need a 64-bit backend to run there
2014-11-14T03:34:13Z nyef: Octeon is... MIPS64?
2014-11-14T03:34:26Z joshe: openbsd doesn't support 32-bit binaries on any 64-bit platform
2014-11-14T03:34:36Z nyef: Ah, damn. That's a no-go for me, then. /-:
2014-11-14T03:34:48Z nyef: See, this is one of the reasons I vastly prefer Linux.
2014-11-14T03:35:02Z nyef: Linux tends to support multilib fairly well.
2014-11-14T03:35:48Z joshe: iirc these octeon boxes come with something debian-based on them
2014-11-14T03:36:10Z joshe: http://www.ubnt.com/edgemax/edgerouter-lite/
2014-11-14T03:38:00Z nyef: Hrm. ISTR looking at these at one point and not being overly impressed.
2014-11-14T03:38:31Z joshe: it's hard to beat $100 for a mips machine
2014-11-14T03:39:18Z christoph_debian quit (Ping timeout: 250 seconds)
2014-11-14T03:39:36Z joshe: well, at least from the perspective of someone who only cares about booting openbsd ;)
2014-11-14T03:42:20Z nyef: Hrm. Half a gig of RAM, no real internal storage...
2014-11-14T03:43:27Z jlarocco quit (Read error: Connection reset by peer)
2014-11-14T03:43:53Z jlarocco joined #sbcl
2014-11-14T03:44:13Z redline6561 quit (Ping timeout: 265 seconds)
2014-11-14T03:44:43Z minion quit (Disconnected by services)
2014-11-14T03:44:46Z minion joined #sbcl
2014-11-14T03:44:58Z nyef: Might be workable, at least.
2014-11-14T03:45:11Z antoszka quit (Ping timeout: 255 seconds)
2014-11-14T03:45:18Z nyef: On the other hand, that'd leave me with an O350 and no real use for it. /-:
2014-11-14T03:45:48Z nyef chants to himself "remember the sunk cost fallacy, remember the sunk cost fallacy".
2014-11-14T03:46:47Z antoszka joined #sbcl
2014-11-14T03:48:55Z redline6561 joined #sbcl
2014-11-14T03:52:03Z nyef: Oh well, even before trying anything new with the O350 or my HPPA or considering getting an edgerouter, I've got a couple of things that need putting together in terms of PPC and SPARC.
2014-11-14T03:52:17Z christoph_debian joined #sbcl
2014-11-14T03:53:55Z joshe: hm
2014-11-14T03:55:42Z nyef: Okay, really I only need to do the SPARC one, and then I can let stassats sort out the rest. (-:
2014-11-14T03:56:02Z joshe: how hard would a 64-bit sparc backend be, do you think?
2014-11-14T03:56:57Z nyef: I don't rightly know, just as I don't know if it could be done as a build option on the current sparc backend or if it'd need to be a full port.
2014-11-14T03:57:48Z nyef: I'd consider a MIPS64 port, myself, but only after getting the basic MIPS backend up to scratch.
2014-11-14T03:58:10Z scymtym_ quit (Ping timeout: 256 seconds)
2014-11-14T03:59:51Z joshe: but I suppose either one is far beyond what I could motivate myself to learn
2014-11-14T04:00:55Z nyef: You might take it as a long-term project. Consider how long the ARM port took, in terms of off-and-on hacking.
2014-11-14T04:02:10Z joshe: I don't think I can find the time and motivation to focus on something like that
2014-11-14T04:03:12Z nyef: I usually can't for more than a few weeks at a time.
2014-11-14T04:03:36Z joshe: when I first picked up a copy of PCL and discovered that there was no (working) openbsd port of sbcl, I just put my head down and spent a month messing around and learning until I got it working
2014-11-14T04:04:20Z joshe: nowadays I'm lucky if I get a solid half day of hacking in on a weekend
2014-11-14T04:05:48Z nyef: I feel the same way at times. Especially weekends, they get to be crazy with other stuff. /-:
2014-11-14T04:06:26Z nyef: You may easily have it worse than I do, of course.
2014-11-14T04:09:52Z joshe: well, my problem is mostly motivation
2014-11-14T04:10:25Z Bicyclidine joined #sbcl
2014-11-14T04:10:43Z joshe: I still haven't finished the kernel hardfloat stuff for arm that I started this summer, not to mention a working arm sbcl port
2014-11-14T04:11:02Z echo-area quit (Read error: Connection reset by peer)
2014-11-14T04:11:48Z joshe: for the last month, I've been trying very hard to work myself up to just write a buildbot config to build sbcl
2014-11-14T04:12:15Z echo-area joined #sbcl
2014-11-14T04:13:50Z nyef: I've been having a hard time recently doing anything in terms of SBCL hacking. The past couple of days have been unusually productive for me.
2014-11-14T04:14:24Z nyef: And, at that, it's mostly forward-porting something I did years ago and filing off the worst of the rough edges.
2014-11-14T04:14:51Z joshe: hm
2014-11-14T04:15:06Z joshe: I stopped working on sbcl right after I got commit access
2014-11-14T04:15:21Z nyef: I go through phases, really.
2014-11-14T04:15:39Z Bicyclidine quit (Ping timeout: 245 seconds)
2014-11-14T04:15:41Z joshe: oh, I need to figure out why ppc threading doesn't work
2014-11-14T04:15:54Z nyef: Up to a month or so of high productivity, then a few months of pretty much nothing.
2014-11-14T04:16:08Z joshe: and commit openbsd-side stuff so i386 threading can be cleaner
2014-11-14T04:16:33Z nyef: I'd say that PPC threading works for me, but I'm no longer so sure that it does.
2014-11-14T04:17:03Z joshe: perhaps ppc is a lost cause
2014-11-14T04:17:34Z joshe: openbsd doesn't even have floating point exceptions on ppc, and multiprocessor kernels apparently aren't very stable
2014-11-14T04:17:41Z nyef: Ouch.
2014-11-14T04:19:08Z nyef: Yeah, I remember digging into the openbsd ppc fp exceptions thing with you.
2014-11-14T04:19:54Z nyef: This would have been back when my main system actually was a PPC Linux box.
2014-11-14T05:04:56Z nyef quit (Quit: G'night all)
2014-11-14T06:00:10Z pranavrc joined #sbcl
2014-11-14T06:00:26Z pranavrc quit (Changing host)
2014-11-14T06:00:26Z pranavrc joined #sbcl
2014-11-14T06:16:03Z gingerale quit (Ping timeout: 244 seconds)
2014-11-14T07:11:53Z oleo__ quit (Quit: Verlassend)
2014-11-14T07:23:31Z kanru` quit (Ping timeout: 244 seconds)
2014-11-14T07:25:42Z kanru` joined #sbcl
2014-11-14T07:35:19Z edgar-rft joined #sbcl
2014-11-14T08:14:11Z fridim_ joined #sbcl
2014-11-14T08:20:04Z White__Flame joined #sbcl
2014-11-14T08:20:09Z White_Flame quit (Read error: Connection reset by peer)
2014-11-14T08:32:13Z pranavrc quit (Remote host closed the connection)
2014-11-14T08:53:57Z prxq joined #sbcl
2014-11-14T09:33:11Z pranavrc joined #sbcl
2014-11-14T09:37:59Z pranavrc quit (Ping timeout: 272 seconds)
2014-11-14T09:56:16Z pranavrc joined #sbcl
2014-11-14T10:07:28Z White_Flame joined #sbcl
2014-11-14T10:07:55Z White__Flame quit (Ping timeout: 250 seconds)
2014-11-14T10:19:24Z White_Flame quit (Ping timeout: 245 seconds)
2014-11-14T10:27:46Z White_Flame joined #sbcl
2014-11-14T10:32:52Z pppp2 joined #sbcl
2014-11-14T10:39:52Z pppp2 quit (Quit: Leaving)
2014-11-14T10:40:18Z echo-area quit (Remote host closed the connection)
2014-11-14T10:53:17Z attila_lendvai joined #sbcl
2014-11-14T10:53:17Z attila_lendvai quit (Changing host)
2014-11-14T10:53:17Z attila_lendvai joined #sbcl
2014-11-14T11:11:42Z pppp2 joined #sbcl
2014-11-14T11:16:31Z pppp2 quit (Read error: Connection reset by peer)
2014-11-14T11:37:04Z attila_lendvai quit (Quit: Leaving.)
2014-11-14T11:43:55Z kanru` quit (Ping timeout: 244 seconds)
2014-11-14T11:45:10Z attila_lendvai joined #sbcl
2014-11-14T11:45:51Z attila_lendvai quit (Client Quit)
2014-11-14T12:33:34Z kanru` joined #sbcl
2014-11-14T12:59:56Z eudoxia joined #sbcl
2014-11-14T13:22:50Z pranavrc quit
2014-11-14T13:35:12Z snuglas joined #sbcl
2014-11-14T13:36:34Z oleo joined #sbcl
2014-11-14T14:08:48Z stassats joined #sbcl
2014-11-14T14:11:23Z stassats: oops, quick-normalization-check is now actually slower than doing normalization
2014-11-14T14:11:23Z minion: stassats, memo from nyef: You might find http://repo.or.cz/w/sbcl/nyef.git/shortlog/refs/heads/read-only-tramps to be interesting. Not ready for "prime time" yet, as it were, but getting there.
2014-11-14T14:15:01Z snuglas quit (Ping timeout: 246 seconds)
2014-11-14T14:40:09Z nyef joined #sbcl
2014-11-14T15:20:43Z DGASAU quit (Remote host closed the connection)
2014-11-14T15:22:39Z DGASAU joined #sbcl
2014-11-14T15:24:47Z stassats: optimizing target-unicode.lisp is fun
2014-11-14T15:25:05Z stassats: probably because it's really unoptimized
2014-11-14T15:25:50Z stassats: now i think i may actually use the normalization feature, for stripping diacritics for searching
2014-11-14T15:47:58Z jlarocco quit (Quit: This computer has gone to sleep)
2014-11-14T15:54:09Z scymtym: scary build failure with ccl host on x86: https://ci.cor-lab.org/job/sbcl-master/738/featureset=7,label=ubuntu_trusty_32bit/console
2014-11-14T15:55:47Z stassats: doesn't seem like sbcl's fault
2014-11-14T15:56:38Z nyef: Does look scary, though.
2014-11-14T15:56:46Z pkhuong: ccl *and* OS X in the same week.
2014-11-14T15:57:13Z stassats: there's also a latent sb-sprof failure, so, there's a lot of things going on
2014-11-14T15:57:25Z scymtym: i have not seen this before, so while likely not being SBCL's fault, it may still have been triggered by a recent commit
2014-11-14T15:57:50Z nyef: And we still have some signal-handling, GC, or allocation issues, don't we?
2014-11-14T15:58:17Z stassats: on some platforms, sure
2014-11-14T15:58:41Z stassats: but the latest known instance should be fixed on x86oids and ppcarmoids
2014-11-14T16:00:08Z nyef: ... ARM is closer to SPARC than to PPC, IIRC.
2014-11-14T16:00:13Z scymtym: there are also sporadic debugger errors on x86: https://ci.cor-lab.org/job/sbcl-master/735/featureset=1,label=ubuntu_trusty_32bit/consoleFull (search for "bad widetag: 38")
2014-11-14T16:00:38Z stassats: nyef: not in terms of the level of sbcl support
2014-11-14T16:01:02Z nyef: Ah, (trace :encapsulate nil)? That's horribly broken anyway.
2014-11-14T16:01:14Z scymtym: well, debug.impure.lisp, not necessarily debugger
2014-11-14T16:01:37Z nyef: Though... Hrm. That's a bizarre failure mode.
2014-11-14T16:01:52Z stassats: scymtym: but, it was at the same line in 731 as well
2014-11-14T16:02:15Z stassats: and 38 is a code object, while it expects to find a fixup vector
2014-11-14T16:02:24Z stassats: inside a code object
2014-11-14T16:02:29Z stassats: so, something goes terribly wrong
2014-11-14T16:02:34Z scymtym: stassats: i never looked at it in any detail, sorry
2014-11-14T16:02:43Z nyef: ... What if that's the fun-end-breakpoint component?
2014-11-14T16:03:18Z stassats: i recently touched trace, maybe i rearranged allocation bits so that it gets triggered
2014-11-14T16:04:23Z nyef: Well, the first place to check is MAKE-BOGUS-LRA in debug-int.
2014-11-14T16:05:01Z nyef: Ugh. Too much x86oid conditionalization. /-:
2014-11-14T16:06:08Z stassats: ok, i guess i need to pummel some allocations with sigprof
2014-11-14T16:07:10Z stassats: how does lra work with x86oids?
2014-11-14T16:07:25Z stassats: just puts it on the return stack?
2014-11-14T16:09:33Z nyef: x86oids don't have LRAs, they use bare instruction pointers. UNALIGNED instruction pointers.
2014-11-14T16:10:04Z stassats: i mean, how does it arrange the return to LRA
2014-11-14T16:10:25Z nyef: Yeah, push it back to the top of the stack and RET.
2014-11-14T16:10:39Z stassats: what is the point of fun-end-breakpoint?
2014-11-14T16:11:47Z attila_lendvai joined #sbcl
2014-11-14T16:11:50Z nyef: You know how some debuggers have a function where you can run the current function until it returns, then reenter the debugger in the calling function?
2014-11-14T16:12:00Z nyef: That's what it's for.
2014-11-14T16:12:08Z nyef: Also useful for trace :encapsulate nil.
2014-11-14T16:12:22Z prxq quit (Quit: No Ping reply in 180 seconds.)
2014-11-14T16:12:29Z stassats: sbcl taught me not to use fancy debugger features, so, i don't know
2014-11-14T16:12:46Z prxq joined #sbcl
2014-11-14T16:12:49Z nyef: That said, SBCL's breakpoint mechanism is horribly dangerous, broken, and that was before it turned out not to be thread-safe either.
2014-11-14T16:13:05Z nyef: Oh, and used by (trace :encapsulate nil).
2014-11-14T16:13:36Z stassats: i need to figure out how trace :encapsulate nil works then
2014-11-14T16:13:53Z stassats: because previously i assumed it modified the code sequence, inserting INT 3 or something
2014-11-14T16:14:33Z nyef: Yes, that's approximately it.
2014-11-14T16:14:44Z nyef: Except that it's not just INT 3, it's a small pile of stuff after it as well.
2014-11-14T16:15:15Z nyef: And then there's the bit where you need to execute a single instruction to get past the breakpoint so that you can replace it, and not all CPUs support single-step...
2014-11-14T16:15:31Z nyef: And then when the CPU has variable-length instructions... Well, you get the idea.
2014-11-14T16:16:04Z stassats: ok, i need to mull over sb-di:make-breakpoint
2014-11-14T16:16:36Z stassats: never touched that part before
2014-11-14T16:18:02Z nyef: Honestly? Just look at arch_do_displaced_inst() in the various backends to see some of the horror.
2014-11-14T16:18:12Z nyef: And then try to figure out how to implement it for ARM.
2014-11-14T16:24:53Z stassats: i did change they layout of code objects, maybe those lras don't mimick it closely enough?
2014-11-14T16:25:27Z nyef: That's also a possibility.
2014-11-14T16:25:44Z nyef: Even a plausibility.
2014-11-14T16:30:17Z stassats: looks more and more like it
2014-11-14T16:31:35Z stassats: wham
2014-11-14T16:31:41Z stassats: i reproduce it
2014-11-14T16:33:57Z nyef: I still haven't managed to figure out if I don't use debuggers very much because they're bug-ridden, user-fiendly pieces of junk, or if they're bug-ridden, user-fiendly pieces of junk because nobody uses them enough to be bothered fixing them.
2014-11-14T16:34:09Z nyef: Which is the cause, and which the effect?
2014-11-14T16:34:40Z stassats: i mostly use them for backtraces and memory poking
2014-11-14T16:34:51Z stassats: break points and steppings rarely works well
2014-11-14T16:34:52Z nyef: If I actually had a good debugger, would I learn how to use it, or would I continue to ignore it?
2014-11-14T16:34:57Z nyef: Yeah, exactly.
2014-11-14T16:35:22Z stassats: now, make-bogus-lra doesn't seem that it would have worked before too
2014-11-14T16:35:26Z nyef: breakpoints and stepping occasionally work well in gdb, especially on non-x86.
2014-11-14T16:35:52Z nyef: What's wrong with it?
2014-11-14T16:36:15Z stassats: it doesn't mind the fixup-vector hiding in the first constant slot on x86
2014-11-14T16:37:21Z snuglas joined #sbcl
2014-11-14T16:41:34Z stassats: i know the fix, but i'm still curious how it worked before
2014-11-14T16:41:48Z stassats: probably the wrong constant numbers just coincided and put things into the right place
2014-11-14T16:42:11Z nyef: Wouldn't surprise me that the entire thing worked by coincidence before.
2014-11-14T16:42:15Z nyef: We've had a lot of that.
2014-11-14T16:42:55Z stassats: building the pre trace-table amputation 32-bit sbcl
2014-11-14T16:43:21Z stassats: but it had the same code for x86-64 and x86, and those can't use the same code, so it clearly was broken
2014-11-14T16:43:24Z nyef: How much do you want to bet that we end up putting the trace-table stuff back in within two years?
2014-11-14T16:43:39Z stassats: i don't care, as long as it is used
2014-11-14T16:43:48Z nyef: Mmm.
2014-11-14T16:43:51Z stassats: but, we can have reliable backtraces in a better fashion, safepoints
2014-11-14T16:44:09Z nyef: Have you taken a look at the read-only-tramps branch that I put together?
2014-11-14T16:44:23Z stassats: partially
2014-11-14T16:45:17Z nyef: I'm thinking to get it running on SPARC and maybe PPC, and file off the worst of the rough edges.
2014-11-14T16:45:33Z stassats: well, i was right, it was broken all this time
2014-11-14T16:45:35Z nyef: Or maybe ARM and PPC.
2014-11-14T16:46:04Z stassats: this suggests, nobody ever used :encapsulate nil on x86
2014-11-14T16:46:44Z nyef: Imagine that.
2014-11-14T16:47:02Z stassats: it works until gc, there's that
2014-11-14T16:48:20Z krzysz00: stassats: Is there any particular reason you removed the "UNICODE1_" prefix on Unicode 1 names (and will anything bad happen if I put it back?) I added that to deal with cases where there are distinct characters X and Y that have (= (name X) (unicode-1-name Y))
2014-11-14T16:48:51Z stassats: krzysz00: there's no point in it
2014-11-14T16:48:57Z krzysz00: One obvious une is U+1F5CF (BELL) and #\Bel
2014-11-14T16:49:41Z krzysz00: With the prefixes, you can say #\USICODE1_BELL for the second character of that pair
2014-11-14T16:50:48Z stassats: nobody will do that or even know to do that
2014-11-14T16:51:23Z krzysz00: Good point. I'll just update the documentation tonight then.
2014-11-14T16:52:55Z stassats: and you asked about what questions target-unicode.lisp raised, well, it looks like it was written intentionally to be slow
2014-11-14T16:53:48Z stassats: but it's moving towards getting faster
2014-11-14T16:56:09Z krzysz00 quit (Ping timeout: 258 seconds)
2014-11-14T17:08:34Z krzysz00 joined #sbcl
2014-11-14T17:19:54Z stassats: ok, encapsulate now works on x86
2014-11-14T17:20:08Z nyef: No, it doesn't. It just doesn't break as obviously.
2014-11-14T17:21:00Z stassats: did i say that i love the regrsssion tests done scymtym?
2014-11-14T17:21:01Z stassats: by
2014-11-14T17:24:43Z eudoxia quit (Quit: leaving)
2014-11-14T17:27:19Z stassats: but i didn't actually start understanding how breakpoints work
2014-11-14T17:27:28Z stassats: since the failure was obvious without that
2014-11-14T17:27:37Z stassats: but it looks like it isn't hard to add to ARM
2014-11-14T17:29:35Z fridim_ quit (Ping timeout: 255 seconds)
2014-11-14T17:29:48Z nyef: In order to be able to use breakpoints, you need arch_do_displaced_inst().
2014-11-14T17:30:32Z Krystof: stassats: quick-normalization-check shouldn't cons, whereas normalization will
2014-11-14T17:31:23Z stassats: i know, but it's really really slow
2014-11-14T17:31:39Z stassats: so, even gcing all the garbage will be faster
2014-11-14T17:32:12Z stassats: i have a way to make decompose-string not cons if there's nothing to do
2014-11-14T17:32:43Z nyef: And since ARM doesn't seem to have a usable single-step, and the instruction set allows using the program counter as a destination for most operations...
2014-11-14T17:34:06Z stassats: and when quick-normalization-check finds that it has to normalize anyway, it's becomes even slower
2014-11-14T17:35:09Z stassats: but ensuring no consing is done should be pretty trivial, just don't collect the chars until there was actually something done
2014-11-14T17:37:05Z krzysz00: Is there anything in particular that's making quick-normalization-check slow? Or is it just a general issue with the algorithm?
2014-11-14T17:37:34Z stassats: ordered-ranges-member is slow, but even removing it, it's still slow
2014-11-14T17:40:12Z DGASAU quit (Remote host closed the connection)
2014-11-14T17:41:37Z krzysz00: ordered-ranges-member could probably become non-recursive if that helps
2014-11-14T17:41:57Z DGASAU joined #sbcl
2014-11-14T17:52:19Z gingerale joined #sbcl
2014-11-14T17:58:05Z krzysz00 quit (Ping timeout: 264 seconds)
2014-11-14T17:58:17Z stassats: ok, got a non-consing and lower consing when the prefix is normalized version
2014-11-14T17:58:58Z stassats: it's slightly slower, but not significantly
2014-11-14T17:59:36Z stassats: still 10 times faster than previously
2014-11-14T18:00:12Z krzysz00 joined #sbcl
2014-11-14T18:01:33Z stassats: can think of even a better way, passing different callbacks to decompose-char, instead of having a callback with two parameters
2014-11-14T18:11:47Z stassats: nice, i like it
2014-11-14T18:31:14Z krzysz00 quit (Ping timeout: 255 seconds)
2014-11-14T18:44:31Z snuglas quit (Ping timeout: 246 seconds)
2014-11-14T18:48:26Z krzysz00 joined #sbcl
2014-11-14T18:48:30Z krzysz00_ joined #sbcl
2014-11-14T18:48:42Z krzysz00 quit (Client Quit)
2014-11-14T18:48:42Z krzysz00_ quit (Client Quit)
2014-11-14T18:49:17Z krzysz00 joined #sbcl
2014-11-14T19:03:48Z krzysz00 quit (Ping timeout: 258 seconds)
2014-11-14T19:11:33Z krzysz00 joined #sbcl
2014-11-14T19:31:41Z milosn quit (Ping timeout: 264 seconds)
2014-11-14T19:32:19Z milosn joined #sbcl
2014-11-14T19:49:47Z krzysz00 quit (Ping timeout: 272 seconds)
2014-11-14T20:08:12Z DGASAU quit (Remote host closed the connection)
2014-11-14T20:10:40Z DGASAU joined #sbcl
2014-11-14T20:17:28Z Bicyclidine joined #sbcl
2014-11-14T20:23:17Z attila_lendvai quit (Quit: Leaving.)
2014-11-14T20:27:06Z cracauer quit (Remote host closed the connection)
2014-11-14T20:27:25Z LiamH joined #sbcl
2014-11-14T20:51:12Z krzysz00 joined #sbcl
2014-11-14T20:54:29Z flip214: christoph_debian: any idea if/when sbcl will land in the ARMEL arch?
2014-11-14T21:02:32Z attila_lendvai joined #sbcl
2014-11-14T21:02:32Z attila_lendvai quit (Changing host)
2014-11-14T21:02:32Z attila_lendvai joined #sbcl
2014-11-14T21:03:39Z stassats: have a test case for signal induced breakage, at least for safepoints-strictly
2014-11-14T21:03:52Z Bicyclidine quit (Ping timeout: 265 seconds)
2014-11-14T21:08:14Z stassats: and the old deadlock when exiting threads otherwise, i think
2014-11-14T21:20:57Z stassats quit (Ping timeout: 240 seconds)
2014-11-14T21:41:42Z Bicyclidine joined #sbcl
2014-11-14T21:47:05Z attila_lendvai quit (Quit: Leaving.)
2014-11-14T21:47:26Z prxq_ joined #sbcl
2014-11-14T21:50:16Z prxq quit (Ping timeout: 265 seconds)
2014-11-14T21:54:57Z jlarocco_work quit (Quit: Leaving)
2014-11-14T21:56:44Z prxq_ quit (Remote host closed the connection)
2014-11-14T22:09:47Z scymtym_ joined #sbcl
2014-11-14T22:56:45Z csziacobus joined #sbcl
2014-11-14T23:00:14Z Bicyclidine quit (Ping timeout: 244 seconds)
2014-11-14T23:53:29Z edgar-rft quit (Quit: end of life during parsing)