00:16:58 -!- Qworkescence [~quad@unaffiliated/quadrescence] has quit [Quit: Leaving] 00:57:51 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Ping timeout: 248 seconds] 01:41:54 borkman` [~user@68.147.21.90] has joined #sbcl 01:43:30 -!- borkman [~user@S0106001111de1fc8.cg.shawcable.net] has quit [Ping timeout: 260 seconds] 01:50:30 -!- borkman` [~user@68.147.21.90] has quit [Ping timeout: 260 seconds] 01:59:59 -!- pchrist [~spirit@gentoo/developer/pchrist] has quit [Quit: leaving] 02:36:52 drl [~lat@110.139.229.172] has joined #sbcl 02:53:14 -!- drl [~lat@110.139.229.172] has quit [Quit: Leaving] 04:43:09 -!- ignotus [~ignotus@unaffiliated/ignotus] has quit [Quit: Coyote finally caught me] 06:08:35 sdemarre [~serge@91.176.68.211] has joined #sbcl 07:09:18 askatasuna [~askatasun@190.97.54.87] has joined #sbcl 07:12:47 -!- sdemarre [~serge@91.176.68.211] has quit [Read error: Connection reset by peer] 07:29:10 sdemarre [~serge@91.176.1.208] has joined #sbcl 08:33:01 attila_lendvai [~attila_le@catv-80-98-25-142.catv.broadband.hu] has joined #sbcl 08:33:01 -!- attila_lendvai [~attila_le@catv-80-98-25-142.catv.broadband.hu] has quit [Changing host] 08:33:01 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl 09:30:37 superjudge [~superjudg@c83-250-198-227.bredband.comhem.se] has joined #sbcl 09:41:23 akovalen` [~user@95.73.218.71] has joined #sbcl 09:43:01 -!- akovalenko [~user@95.73.109.170] has quit [Ping timeout: 240 seconds] 09:44:41 -!- askatasuna [~askatasun@190.97.54.87] has quit [Quit: WeeChat 0.3.5] 11:20:43 homie [~levgue@xdsl-78-35-160-109.netcologne.de] has joined #sbcl 11:21:00 -!- superjudge [~superjudg@c83-250-198-227.bredband.comhem.se] has quit [Quit: superjudge] 11:21:26 hargettp [~hargettp@pool-71-184-177-139.bstnma.east.verizon.net] has joined #sbcl 11:38:08 kennyd [~kennyd@93-138-109-254.adsl.net.t-com.hr] has joined #sbcl 11:39:09 superjudge [~superjudg@c83-250-198-227.bredband.comhem.se] has joined #sbcl 11:42:05 -!- superjudge [~superjudg@c83-250-198-227.bredband.comhem.se] has quit [Client Quit] 12:20:38 -!- akovalen` is now known as akovalenko 12:46:17 -!- kennyd [~kennyd@93-138-109-254.adsl.net.t-com.hr] has quit [Disconnected by services] 13:50:57 -!- Krystof [~user@81.174.155.115] has quit [Ping timeout: 252 seconds] 13:59:18 -!- antifuchs [~foobar@care.boinkor.net] has quit [Ping timeout: 244 seconds] 14:01:39 antifuchs [~foobar@care.boinkor.net] has joined #sbcl 15:21:14 -!- hargettp [~hargettp@pool-71-184-177-139.bstnma.east.verizon.net] has quit [Quit: Leaving...] 15:37:15 Krystof [~user@81.174.155.115] has joined #sbcl 15:48:46 nyef [~nyef@c-174-63-105-188.hsd1.ma.comcast.net] has joined #sbcl 15:48:59 Hello all. 16:07:20 Okay, after assessing my two main SBCL trees, I can conclude that my major outstanding branches are x86oid disassembler improvements, setf cleanups, and wider-fixnums. Maybe another branch somewhere, too. 16:21:30 not bad 16:21:52 I have work-in-progress on rational as read-default-float-format and subclassable structures 16:22:19 as well as an initial stab at array type cleverness 16:22:20 Krystof: rational as read-default-float-format? but that is sooo non-portable! 16:22:25 *akovalenko* just couldn't resist 16:22:38 akovalenko: It's called "stealth lock-in". :-P 16:23:06 it wouldn't be the default 16:23:27 Oh, and I remembered approximately why I didn't commit wider-fixnums, and was completely unable to verify that it was any sort of a problem. 16:24:09 I don't mind unportable options: otherwise I wouldn't have messed around with sequence subclassing 16:25:09 Is the september release still being held up for anything? 16:26:19 Hrm. All of my git trees still point to boinkor as origin. 16:33:52 I'm just going to revert the bogus nanosleep test and maybe update the sleep consing test to be an expected failure, and then release 16:34:31 I'm having trouble finding this, so... what's the read-write clone URL? 16:36:03 ssh://SFUSER@sbcl.git.sourceforge.net/gitroot/sbcl/sbcl.git 16:36:08 Ah, thank you. 16:37:08 Yay! My ssh key is still good. I have /one/ way to access my sourceforge account. 16:42:39 *nyef* sighs. 16:42:46 "./make.sh: 233: time: not found". 16:47:16 quick question: (speed 0) (debug 3) (safety 3) doesn't do tail-call optimization. What's the safest way to enable that feature while staying as safe as possible? 16:48:09 (need TCO for infinite recursion) 16:48:48 Phoodus: Not entirely certain, but... possibly a so-called "dependent quality" in optimize will do the trick. 16:49:14 yeah, just asking if somebody here has that fact in head-space at the moment ;) 16:49:50 Possibly sb-c::merge-tail-calls, but I'm really not sure. 16:52:30 Phoodus: ISTR space is important (probably, (space 2)) 16:53:19 I've been thinking for a while that it might make sense for the compiler to /never/ make a decision on the basic qualities and only work in terms of dependent qualities. 16:54:12 I agree, for the general intent of config. Though having very feature-specific flags for debugging is also nice ;) 16:55:45 It would make the actual effects of the basic qualities easier to predict, make the intent of any policy check more obvious, and make it easier for users to determine how to get a specific effect more easily. 16:56:24 In theory, at least. 16:56:32 Not sure how it'd all work out in practice. 16:57:49 via google, nikodemus says if (> SPACE DEBUG) then it's enabled. I dropped debug to 2 and it seems to be TCOing 17:05:05 hmm, bytes-consed-between-gcs is still 32-bit in the sourceforge browser 17:05:31 I never got around to submitting that patch, should do so unless one of you guys wants to flip that over :-P 17:06:30 heh, it's word-sized in C code, but 32-bit in gc.lisp :) 17:07:48 yep, I don't seem to have my local changes for that anymore; can you fix that? 17:08:23 *akovalenko* can't fix anything in the mainline SBCL, though sometimes has a chance to spot something.. 17:21:17 akovalenko: Are there things in your branch that could be usefully merged without the rest of your changes? 17:23:55 nyef: yep (e.g. long file name support -- I backported it into mainline-based branch myself). 17:25:44 nyef: let's wait for _8david's thread/safepoint commits, though (there'll be a lot of stuff in them, and I won't dare to touch anything that can intersect with them) 17:30:17 Mmm. What about the isatty() on win32 thing? 17:30:51 Are we moving entirely to a safepoint model for asynchronous interrupts? 17:32:08 And I seem to recall something about PROBE-FILE on UNC filespec shares, did that ever get fixed? 17:34:03 isatty() ;; it's really GetFileType() == FILE-TYPE-CHAR here, now 17:34:54 interrupts ;; on windows, we're moving from no interrupts at all to safepoint-based one. Ask David Lichteblau for what he plans for POSIX. 17:36:43 probe-file on UNC ;; used to work for me due to an ad-hoc kludge in query-file-system, and now it works as a side-effect of LFN-awareness (I've abandoned all MSVCRT file api altogether) 17:37:25 Mmm. There's something dangerously stupid surrounding POSIXoid interrupt handling... Something like most signals should only be added to the deferred set when there's a lisp handler for them, and should be removed from the deferred set when the lisp handler is removed. 17:38:02 nyef: async lisp handlers work by accident, end of story (that's how I see it). 17:38:40 nyef: so moving it all to safepoints is inevitable, the question is "when". 17:38:53 The sooner, the better. 17:39:28 And then WITHOUT-INTERRUPTS can help by not emitting safepoints within its body... 17:39:52 Or something like that. 17:39:53 nyef: without-interrupts is not lexical-only 17:40:03 True, but it'd speed up the lexical cases. 17:40:57 I'm with you both. _8david differs, iirc. 17:41:05 The real trick would be to have W-I handled via out-of-band code annotation, like DWARF unwind info. 17:41:51 So, how many people would be inconvenienced by a shift to 63-bit fixnums on x86-64? 17:42:02 currently, safepoint is one memory-load instruction from a fixed location that should be almost always in cache. I'm unsure if it's possible to notice the difference, though I've not measured it yet 17:42:48 nyef: well... I suppose that in-L1 simple-vector accesses would be slowed down a tiny bit. 17:43:19 akovalenko: perf-wise? Noise, around 1-2%. 17:43:35 nyef: Many users will be inconvenienced by the bugs caused by the assumption of word-shift == fixnum-tag-bits. 17:43:47 pkhuong: And if that inconveniences anyone, it'll turn out that they maintain their own branch anyway, and can easily switch it back. 17:44:12 akovalenko: And if either symbol were exported from a public package, I'd consider shedding a tear for them. 17:44:39 akovalenko: I don't think anyone except sbcl devs play that sort of tricks with g-l-o-a. 17:45:17 yep, and I didn't mean third-party bugs :) 17:45:36 akovalenko: nyef's branch includes fixes. 17:46:01 Mmm. One of the patches in the series is entirely about sorting that mess out. 17:47:21 I guess the follow-on question is, should we try and get 31-bit fixnums on non-x86-64 targets? :-D 17:48:28 hmm, if only sbcl was not so good in unboxed arithmetic.. actually, /me never missed wider fixnums in SBCL for this reason :) 17:48:31 It only requires doubling the alignment requirement for boxed data, which would surely cause problems with using SIMD instructions on arrays by forcing their data vectors to that unaligned offset... 17:51:22 (Okay, doubling the alignment requirement, and I think it requires losing a lowtag as well, making some type-tests slightly more expensive.) 17:51:41 nyef: doubling alignment requirement can only help SIMD. 17:52:27 pkhuong: Yeah, I know. You just burn a slot, and the data vector then always starts in the right place instead of the wrong place. 17:53:10 ... unsynchronized hash-table tests are disabled on x86-64/linux? 17:54:53 nyef: disabled period. It sometimes fails ungracefully. 17:55:44 Ah. 18:06:36 Okay, so I have three expected failures, one failure (signals.impure.lisp / SLEEP-MANY-INTERRUPTS), and seven skipped. Is that about right? 18:06:52 sleep many is bogus. That sounds right. 18:26:59 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Read error: Operation timed out] 18:56:18 superjudge [~superjudg@c83-250-198-227.bredband.comhem.se] has joined #sbcl 19:18:33 -!- superjudge [~superjudg@c83-250-198-227.bredband.comhem.se] has quit [Quit: superjudge] 20:19:31 hargettp [~hargettp@pool-71-184-177-139.bstnma.east.verizon.net] has joined #sbcl 20:30:07 -!- sdemarre [~serge@91.176.1.208] has quit [Ping timeout: 248 seconds] 20:59:46 attila_lendvai [~attila_le@catv-80-98-25-142.catv.broadband.hu] has joined #sbcl 20:59:46 -!- attila_lendvai [~attila_le@catv-80-98-25-142.catv.broadband.hu] has quit [Changing host] 20:59:46 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl 21:13:49 -!- attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has quit [Ping timeout: 240 seconds] 21:18:40 -!- hargettp [~hargettp@pool-71-184-177-139.bstnma.east.verizon.net] has quit [Quit: Leaving...] 21:20:25 attila_lendvai [~attila_le@unaffiliated/attila-lendvai/x-3126965] has joined #sbcl 21:28:01 hargettp [~hargettp@pool-71-184-177-139.bstnma.east.verizon.net] has joined #sbcl 21:30:33 -!- hargettp [~hargettp@pool-71-184-177-139.bstnma.east.verizon.net] has quit [Client Quit] 21:37:13 Did someone measured how fast is FAST-READ-CHAR (&friends), relative to some pre-optimized version? (e.g. it's sometimes useful to load fasls from Gray streams, but direct uses of FAST-READ-BYTE prevent it..) 21:38:33 You should be able to create an ANSI-STREAM wrapper around gray streams for the benefit of the fasloader. 21:39:15 At the same time, however, I have a couple of ideas for... substantial damage to the FASL format. 21:40:04 heh, will something like (make-concatenated-stream) be almost sufficient? (somehow it's required to be a file-stream as well, but that place is not essential) 21:40:32 I have no idea. 21:41:10 Umm... Remember that you can blast the misc-routine of an ansi stream, and also other fields, if you need to cheat to get something through the system. 21:42:47 I have at least once overridden the misc-routine of a stream for some effect, but I forget the actual context. 21:43:11 yes, I can (I'm also doing something like that, getting read-only file streams to be mmapped while they stay ANSI; SBCL believes that the whole content is buffered, etc) .. 21:43:32 Heh. Nice! 21:43:39 *akovalenko* had in mind something about normal SBCL users.. 21:49:46 homie` [~levgue@xdsl-78-35-149-137.netcologne.de] has joined #sbcl 21:52:34 -!- homie [~levgue@xdsl-78-35-160-109.netcologne.de] has quit [Ping timeout: 244 seconds] 22:05:02 -!- homie` [~levgue@xdsl-78-35-149-137.netcologne.de] has quit [Quit: ERC Version 5.3 (IRC client for Emacs)] 22:07:53 homie [~levgue@xdsl-78-35-149-137.netcologne.de] has joined #sbcl 22:56:41 -!- nyef [~nyef@c-174-63-105-188.hsd1.ma.comcast.net] has quit [Quit: G'night all.] 23:38:39 -!- Quadrescence [~quad@unaffiliated/quadrescence] has quit [Quit: bonk] 23:40:01 Quadrescence [~quad@71-208-99-228.hlrn.qwest.net] has joined #sbcl 23:40:04 -!- Quadrescence [~quad@71-208-99-228.hlrn.qwest.net] has quit [Changing host] 23:40:04 Quadrescence [~quad@unaffiliated/quadrescence] has joined #sbcl