| |
| |
| |
|
Page: 1 2 3 4 5 6 7 8
Comments:
<0> (Asking as I don't see Nyef around) anybody on know anything about Win32 SBCL?
<1> what is "Attempt to access unrecorded object (id 12)." when tryin to evaluate a defun?
<1> oh seems to be somthing wrong with having \#Soh in code
<1> nothing I (format t ..) is going to anywhere I can see
<1> why? :(
<1> well, its going ot the wrong plac
<2> nothing to stress a garbage collector like ... another garbage collector
<3> slyrus:
<2> antifuchs:
<3> yeah, RET instead of backspace syndrome
<3> (the usual) (:
<3> slyrus: anyway, I wanted to ask if you are implementing a separate gc (:
<2> antifuchs: rucksack has a persistent-storage gc
<3> ooh
<3> you're hacking on rucksack? interesting!
<2> it does some neat things: 1) it stresses the GC and occasionally exhausts the heap and 2) lately, now that i've managed to avoid 1), gives me lots of fake_foreign_calls falling through.
<2> fun fun fun
<2> there's about 75 emails on the rucksack list from the last day and a half describing my experiences
<2> I'm trying to cram 1M objects into rucksack. rucksack and/or SBCL don't really appreciate the effort.
<4> glhs format string
<4> clhs format string
<2> clhs format
<5> http://www.lispworks.com/reference/HyperSpec/Body/f_format.htm
<3> slyrus: well put (:
<2> it's kinda fun, and it will be neat if it works, but I'm skeptical
<4> slyrus: why it is so damn difficult in the clhs to get to the page which explains the structure and functioning of the format control string?
<6> it's not so damn difficult
<3> just remember the section number and you're set (:
<3> also, slime has help for ~ escapes in format strings
<4> :)
<3> C-c C-d ~, iirc
<4> jsnell: i don't manage to find it (again) :-(
<6> clhs 22.3
<5> http://www.lispworks.com/reference/HyperSpec/Body/22_c.htm
<3> valizas: you are using slime, right? (:
<4> ok 22.3 is my number of luck :) thankyou
<4> antifuchs: not yet
<6> some ways on how to find it: go to clhs page for format, click on the "For details on how the control-string is interpreted, see Section 22.3 (Formatted Output)." link
<4> just emacs+cmucl+my browser :)
<3> valizas: you should! it has some very fine functions to help you hack lisp
<6> or go to front page, click on "Printer", than on "Formatted Output"
<6> neither of these seem particularily hard or unintuitive to me
<4> I think i will help to get slime up and running, would you ?
<4> s/will/will need/
<3> it definitely will. it has hyperspec lookup bound to a similar key sequence, by the way
<3> oh
<7> valizas: If you don't use Slime, why use Emacs at all? Hemlock is way cooler. :)
<4> too :)
<3> no, should be pretty straightforward
<6> unlike, say, 11.1.2.1.2.1 which is really easiest to find by remembering the section number
<4> rydis: you you want to laugh while i tell the story ? :)
<3> download, add to load path, add a setup statement to your .emacs, M-x slime
<4> antifuchs: ok tks
<2> antifuchs: two control keys and a grapheme, what could be easier? :)
<7> valizas: Ignore me. Listen to the helpful people, instead.
<3> slyrus: I still remember gnus message mode gpg signing shortcuts
<4> rydis: what you signal is cool too.
<3> C-c RET s for sign or something... ew
<3> or was it C-c TAB? something horrible
<4> either hemlock or slime will be better than just emacs loading cmucl
<4> but beforehand it was worse. I called cmucl from the prompt and hacked in there, no line editing, no nothing
<3> I have a feeling you'll enjoy slime (:
<4> antifuchs: i agree, i'll enjoy every little step that glorifies my work and learning
<3> (:
<1> what is hemlock like?
<3> bitter.
<2> tasty, but you feel like hell in the morning
<3> (scnr)
<3> slyrus: ^5 (:
<7> fax: Hemlock is like somewhere between GNU Emacs and TECO EMACS.
<7> (With a smattering of ZWEI, possibly.)
<7> It also has absolutely no support for modern X stuff, like colours (OK, it has, actually, but only two), selections, non-ascii characters, or any of that crap. :)
<2> antifuchs: have you used rucksack?
<3> slyrus: no, only heard the talk in hamburg
<2> ok. I think it needs more users/developers. it's a neat idea and seems to work for toy examples, but it's going to need a little love before it scales to big data sets, IYAM.
<3> ic... how large datasets are you thinking about?
<3> (the previous sentence was not entirely grammarified)
<2> oh, ATM I've got lots of small obejcts. around 1M objects in total.
<3> ah, that doesn't sound too uncommon
<2> maybe a bit more.
<2> yeah, it's not like I'm trying to index 100M objects
<3> I wonder... the bknr guys had lots more data around, at least in the eboy website (iirc)
<3> but they didn't have a gc, did they
<2> the ironic thing is that the data set I have now is small enough to fit in the heap, if you don't cons up all the extra crap for the persistent storage :)
<3> haha
<3> you should just mmap the heap to a rw file and be done with it (:
<2> yeah, jsnell suggested something like that.
<2> I get the impression that I'm one of the first folks to push rucksack this hard. I don't think it's hopeless, but it certainly needs folks to exercise it (and to fix the resuliting problems)
<2> I'd like to combine rucksack with spatial-trees for spatial indexing and, perhaps, montezuma for text indexing.
<3> yum
<2> well, if it works anyway :)
<8> 2d spatial indexing, or higher dimensional?
<2> 2d to start
<8> I don't anticipate things in the R tree family scaling well to much higher than 4d
<8> if you want to go higher, then LSH might be the way to go
<2> yeah, agreed.
<3> good night (:
<9> antifuchs: goodnight
<9> where do i have to untar slime ?
<7> valiza: You should probably get slime from CVS.
<7> (And the answer to your question is "somewhere, and put it in Emacs' load-path".)
<9> tks !!
<9> is the cvs version better?
<7> The releases are very infrequent.
<9> ah ok
<9> i think i made emacs slime-aware. How do i start slime?
<10> M-x slime
<9> gosh
<9> woooooooooowooooooooooooo
<9> this is too much for me
<11> hrm, anyone a prevalence proponent?
<12> Krystof: LSH?
<9> oGMo: Wuestefeld :-)
<11> valiza1: heh
<13> sbcl on osx/intel crashes on me on startup: The application sbcl quit unexpectedly ... EXC_BAD_ACCESS ... KERN_PROTECTION_FAILURE ...but then i can still run code.
<9> don't know how far Klaus W. has gone with Prevayler
<11> i'm not clear on the perspective of various fundamental things, like ... crash recovery
<9> simple you take the last snapshot, and apply the log since then to get the last logged state
<11> oh, so stuff is actually written back to the filesystem as you go, just kept in ram. wasn't entirely clear. that works for me.
<11> oh, duh, how did i miss this ;P
<14> oudeis__, I believe this is expected, because of the way SBCL triggers garbage collection.
<2> oudeis__: it crashes or it tells CrashReporter that it crashed?
<9> the main principle is: nothing changes in ram until it got written to the log
<14> ...oh, ignore me; someone who knows what he's talking about is here now.
<13> slyrus, exactly
<13> crash reporter reports this
<2> oudeis__: you have a couple options: turn off CrashReporter or use the mach-exception-handler feature and rebuild
<2> nyef and I rigged up SBCL to use mach exceptions instead of unix signals for some of the signalling stuff and this fixes the crashreporter problem, but it is not on by default.
<13> slyrus, thanks. does it have any adverse effects?
<2> I think lemonodor has some blog posts that describe the symptom and workarounds.
<2> well, the most adverse effect is that it seems to mess up threads even more than usual. if you're running unithreaded, you probably won't care
<2> and it isn't particularly well tested ATM
<2> there's another benefit which is that you can run SBCL under gdb with the mach exception handler version, which is nice.
<13> i do run threaded unfortunately ...
<13> but only on development machine
<2> well, then turn off CrashReporter, or turn on mach exceptions and threads and fix the remaining threads bugs
<13> :)
<13> how often would you expect it to mess with threads?
<13> is it bearable or not?
<2> try it and see. it was for a while, then it wasn't. not sure what's going on.
<13> ok. thanks!
<10> its bearable if you develop on it and plan to deploy on linux
<10> crashes are not very frequent
<13> xristos, so you run it?
<10> yeah
<13> ok
<10> i don't do heavy multithreading though
<10> no locks etc
<13> tell Riastradh thank you too, why ignore a good piece of information :)
<2> yay. acceptable performance on 750k objects in rucksack.
Return to
#lisp
or
Go to some related
logs:
#css
+linux +crond +multiple instances of crond
#web
pixma mp110 can't install
#css
#perl
#php
You do not have the permissions necessary to view the contents hal
ubuntu lexmark z615
#fluxbox
|
|