| |
| |
| |
|
Page: 1 2 3 4 5 6
Comments:
<0> don't forget to increment version.lisp-expr :)
<1> I did that.
<0> it's always interesting to trace the performance effects of BUGS-only commits, as reported by boinkor ;)
<0> (thanks for committing btw; I'm very much the slacker these days wrt sbcl development)
<2> sorry for the OT question: does anyone know how to boot Linux from LILO in single-user mode (I forgot my p***words on this old machine).
<3> beach: add a -s to your boot command line?
<0> something like init=/bin/sh
<4> beach: init=/bin/sh
<1> I trust in the ability of the great interweb to finish the win32 port. I can test and commit it. :-)
<3> Note that some distros will still ask for p***words if you boot -s...
<2> OK, I tried the -s and it asked for an admin p***word, and then I could not log in as root with that p***word.
<3> cliini: Does sb-grovel still not use -mno-cygwin on win32?
<3> beach: RedHat box?
<2> nah, a very old debian
<1> nyef, apparently so, I had to add that.
<2> I'll go try the init=/bin/sh thing, thanks.
<3> Well, try the init= trick. That should work if you're not using an initrd (and maybe even if you are).
<5> beach: my suggestion is to boot off a live CD, mount your disk's filesystems and edit /etc/shadow
<1> I'll have to figure out how to fix it properly, as the place I added it to seems to be generated.
<5> init=/bin/sh will work too but you have to be careful, because you have no job control (so things like C-c to abort something don't work)
<1> Err, no, it's not generated, I just managed to overwrite my change.
<6> thus the first thing you should do is run bash/dash/zsh/tcsh/whatever. ;)
<1> I'll beautify the fix and commit it tomorrow.
<5> jsnell: off the top of your head, can you remember if symbols' hash values are guaranteed the same between lisp sessions?
<7> How do you call lists that are shaped like lambda-list, they ain't proper property lists, so..?
<8> lambda list?
<7> Well, I didn't mean it like that, but nevermind, I just decided to do it differently anyway what I had in mind.
<2> Xof: thanks, that worked, but when at the shell prompt I try to change the p***word, I get an error message; something like token manipulation error. Any clues:
<2> s/:/?/
<5> I wouldn't expect the p***wd binary to work
<2> ah, that would explain it.
<5> instead, edit /etc/shadow directly and delete the root p***word
<2> after booting off a live CD?
<5> you can do it with the init=/bin/sh method too
<5> in both cases you have to mount the filesystem with /etc read-write
<5> usually you can do this with mount -orw,remount /
<5> then you need to edit /etc/shadow
<2> ah, OK. I'll try that.
<5> then you need to mount -oro,remount / again (because otherwise it might not flush the disk cache)
<5> then you should be good to reboot :-)
<0> we'll make a h4x0r out of him, yet :)
<2> thanks a lot
<9> xof, beach: Possibly also /etc/p***wd (depending on the patriculars).
<0> Xof: does executing /bin/sync instead of remounting ro have the same effect?
<3> rudi: No, because /bin/sync doesn't set the bit which indicates a clean unmount.
<5> rudi: it helps. But unless you unmount cleanly, you get to fsck on next reboot
<0> ok, thanks
<5> argh so many bugs
<3> Heh. I was just thinking the same thing about something else.
<2> xof's advice worked fine. Thanks again to everyone. Now I have to get out of here to test my other ADSL line.
<10> Xof: yes
<10> (IIRC, it's actually mandated by the spec)
<11> Xof pasted "ow ow ow" at http://paste.lisp.org/display/24115
<12> that loop form looks malformed to me
<5> why?
<12> but, it's possible I just don't remember which arguments to loop take multiple forms :-)
<5> do takes multiple forms
<5> what's rather more exciting is the failure mode
<13> hrm, when i try to load slime i get an error about can't load 'timer' ?
<13> Cannot open load file: timer
<13> i've tracked it down to being within slime. Is there a module i'm missing or something?
<14> i have the same problem in xemacs
<13> yep
<14> afraid i can't be of much help though as i have no clue about emacs
<13> :(
<13> i'm not usually an emacser either, but it's gotta be the best thing for lisp
<4> You seem to have a version of Emacs that doesn't have timer, which seems odd.
<9> Theoretically, there should be a "timer.el2 or "timer.elc" in a directory taht's in load-path. There's one "here", for my GNU Emacs.
<4> I'm fairly certain it comes with Emacs, but maybe you're on an odd platform?
<13> james@opiate ~/***pr_1.1/ruby $ xemacs --version
<13> XEmacs 21.4 (patch 19) "Constant Variable" [Lucid] (i686-pc-linux) of Mon Aug 7 2006 on opiate
<13> nope, gentoo linux
<4> I don't know anything about XEmacs, but Gentoo has a reputation for being a bit odd.
<13> haha. we aren't that odd. let me see what the useflags are for it
<3> "aren't that odd"? Last I knew, Wine refuses to deal with bug reports if the reporter admits to using the gentoo ebuild.
<13> haha. well that's just bizarre. but then the wine team are
<4> Anyway; timer.el comes with regular Emacs. I don't know about XEmacs. If you're not an XEmacs user, you might want to try with regular Emacs rather than XEmacs.
<3> And that because the wine ebuild is completely different from the way the default wine setup is from upstream.
<15> i have a timer.el for my xemacs (debian package), i didn't install it seperately so it likely came with it.
<4> timer.el seems to come with regular XEmacs, too.
<13> usually they just bend it a little to fit the way gentoo likes things. besides they should be grateful, we're probably responsible for about half bug reports
<14> i have itimer.el{c,} from xemacs but no timer.elc
<3> And you have a number of people here, including a few gentoo users, who recommend not using the ebuilds for anything lispy.
<13> fine, i'll install dpkg
<13> (that means i can get lisp in a box right?)
<3> (And at this point, my recommendation is just to not ever use vendor packages for lisp stuff.)
<13> so much for that idea then
<13> i generally use it because it takes care of all the dependencies nicely
<16> you want lisp in a box?
<13> i'm just downloading it now
<13> it seems to be a convenient way to go about it
<6> the debian one seemed to work well enough for me until I wanted to apply my own patches.
<13> okay, i'll be blunt, what's the best way to get a lisp environment set up if i can't use portage or LIAB ?
<9> On the whole, I've not had that much bad luck with Debian-packaged SBCL. It's not always bleeding-edge, but I seldom require bleeding-edge.
<13> i don't mind a bit of work, what's the recommended way
<9> But prior to that, I donwloaded SBCL-source and built myself.
<9> SBCL is, at base, easy to build (and re-build). Never, really, tried multi-platform cross-compile.
<4> It seems that XEmacs has timer.el in some "fsf-compat" package, which you might need to download separately.
<13> hrm
<4> ElPenguin: Do you have any particular preference for XEmacs?
<13> rydis: not really
<6> ISTM that xemacs is kinda...falling behind.
<6> for a while gnu emacs was dead and lemacs/xemacs was the hot thing. now it seems mostly the other way.
<4> ElPenguin: Then try the package for regular Emacs, download the latest SBCL release, get slime from CVS, and you should be ready to go.
<4> (Or just use CMUCL and Hemlock, like I do. ;)
<13> i did try in regular emacs. no error but no slime either
<17> conservative experiment: http://paste.lisp.org/display/24116
<9> I did actually ponder something this morning. I am beginning to believe that language maturity is measured by home many emacs re-implementations are available.
<9> Common Lisp should, therefore, be fairly mature (Hemlock, Portable Hemlock, Climacs and the Lisp Machine emacsen).
<9> Python-the-language has pymacs. C has gnu-emacs, xemacs and possibly more.
<18> mega1: So you have GC issues too?
<4> ingvar: zile, uemacs, ...
<4> (For C, that is.)
<17> rtoy_: just overconservatism. I think the weak hash tables work well now.
<9> There's perlmacs (or was, possibly) for Perl.
<18> mega1: Yeah, that's what I meant. Cool, though!
<17> rtoy_: it's only x86/x86-64 at the moment but supports :key :value :key-and-value :key-or-value too
<17> testing is next to impossible on gencgc
<5> mega1: you need a ppc :-)
<18> mega1: Oh, that's nice! I thought about adding the extra weak types, but got lazy.
<5> bah. jsnell: can you think of a way of implementing a total order on symbols?
<18> mega1: I ***ume you didn't have to sprinkle scan_weak_tables all over?
<17> rtoy_: not exactly, but I think the cmucl impl is buggy
<5> jsnell: just using symbol-hash doesn't work because there are collisions. (And common ones at that: anything with the same symbol-name)
<10> right
<10> using the address is out, since they can move
<5> I'm looking at symbol-lessp in src/pcl/vector.lisp, thinking "that's so wrong", but everything else is wrong too
<18> mega1: I'm not surprised. Which part is buggy? The weak tables or the overall GC part?
<17> rtoy_: the three places where scan_weak_tables are in scavenge_newspace are imho right, but it shouldn't be scan_weak_tables that's called
<10> so I can't see how to do it without adding a new slot (or using a soon-to-be-working weak hash-table)
<5> maybe the answer then is to rip out this sorting thing
<17> instead one must perform repeated scavenges _without_ removing the unreferenced elements
<17> once scanvenge_newspace is done then remove the unreferenced elements and then break the weak pointers
<17> One must not remove the unreferenced elements before scavanging is complete for the obvious reason that another ref may be found later.
<17> Furthermore, I modified free_hash_entry to look for the entry on the rehash chain too instead of just giving up.
<18> mega1: I see.
<17> I'm struggling with testing immensely.
<18> mega1: I guess I'll wait until you make a commit or send a patch. :-)
<17> that's probably for the best, I may find more bugs before the sun goes down
<18> mega1: Out of curiosity how did you implement the other weak table types? Allow :weak-p to take different values besides T and NIL? How did you chain the weak tables together then?
<18> Or add a new slot for the weak type?
<17> rtoy_: I changed the name to WEAKNESS and allowed :key, :value, etc. And also added a NEXT-WEAK-HASH-TABLE slot to chain them during gc time.
<18> Ok. Makes sense.
<18> mega1: Also, you basically want to scav_weak_entries but without scav_hash_entries, until the newspace has been scavenged? Then at the end, scan_hash_entries to remove the weak entries?
<17> rtoy_: I don't have the cmucl source here at work, and I renamed the function because they were confusing to me :-). Scav_weak_entries doesn't exiss here anymore: it was absorbed by scav_hash_entries, and yes scan_hash_tables removes the dead elements.
<18> So scav_hash_entries basically scavenges the key/value vector but that's all? Which is what scav_weak_entries kind of did, I think. I guess you also have several more functions to handle the different weak table types?
<17> rtoy_: it's a single function, within the keyvector loop the scavenging of entries only happens if ((weakness == NIL) || ((weakness == KEY) && survived_gc_yet(old_key)) || ((weakness == VALUE) && survived_gc_yet(value)) ....
<17> what scav_weak_entries could not do is rehashing. It got away with that since it only supported :key weakness. But for :value that's not enough so that code moved into scav_hash_entries.
<18> Ok. I'm just looking at the code, and adding a next-table slot for chaining and maybe changing the weak-p slot to indicate the weak type, instead of overloading that for chaining. The other stuff will come later.
<19> morning
Return to
#lisp
or
Go to some related
logs:
#web
Mario Mommers
prolificgnosis
cdtodvd
#css
#perl
gentoo install Could not activate swap
linux lide60 suse
blocksshd howto
python walk hidden files
|
|