The Return of the Journal : pugs.blogs.com/ | pugscode.org | pugs.kwiki.org | paste: paste.lisp.org/new/perl6 or sial.org/pbot/perl6
Set by GammaRay on 31 December 2005.
00:12 pdcawley joined 00:20 brentdax joined 00:22 pdcawley joined 00:36 Cryptic_K joined 00:37 Zobbo joined 00:41 athomason joined 00:43 Zobbo left 00:44 xinming joined
dduncan is sigwinch and stuff a new feature in readline? 01:07
01:07 putter joined
dduncan perhaps audreyt and some of you have a newer readline version than I do 01:07
putter hi dduncan. a fresh r8825 compiled for me. checking libreadline... 01:08
dduncan I just reverted my Pugs.Shell to r8824, and it now compiles without trouble 01:10
putter i'm at 4.3, (hmm, and segfaulting :/ ah well - os upgrade coming soon). 01:11
dduncan so the r8825 is making some assumption that isn't true for my system
putter what was one of the undefined symbols?
dduncan backlog by about 3 hours
putter looking...
dduncan see colabti.de/irclogger/irclogger_log/...l=357#l607 01:13
01:13 xinming1983 joined
putter hmm, rl_catch_signals is just an int. is it defined in your /usr/include/readline/readline.h or equivalent? 01:16
the next question will be, is it defined in your libreadline. eg, locate libreadline; then strings /usr/lib/libreadline.4.2.so |grep rl_catch_signals or some such. 01:23
it looks like rl_catch_signals is an 4.2'ism - googling turns up config files which test for it separately... 01:28
exploring further...
dduncan my /usr/include/readline/readline.h doesn't have them
mine is $NetBSD: readline.h,v 1.11 2004/01/17 17:57:40 christos Exp $, copyright 1997 01:29
brentdax Rules are segfaulting right now. Expected or no? 01:30
dduncan from my reading on the 'net, it seems that 10.4 is the first Mac OS X version to bundle readline, but online reports say it is buggy ... also, this library is used by a number of third party apps 01:34
see, for example, www.justatheory.com/computers/datab...dline.html 01:35
avar olofyrla:~ avar$ locate readline.h
olofyrla:~ avar$
you mean bundled with the 10.4 dev tools I presume?
dduncan or that
I have 10.4.4, plus the devtools 2.0 01:36
it says XCode is 2.1, though I don't use that program ... mainly I installed the devtools to get cc and stuff 01:37
avar I love how they make you sign up for their developer network just to brag about them having 10 million developers or something 01:38
..when you just need to dl a C compiler
dduncan in my case, the devtools were bundled with the retail Tiger dve 01:39
dvd
avar It's like those "vote for us to download" w4r3z websites I used to go to before I hit puberty
dduncan I think the devtools were always bundled with Mac OS X, so you don't have to download them
unless you want versions newer than came with the os 01:40
avar or I don't want to go digging around for some CD
dduncan I just install them when I install the OS and be done with it
avar I just throw the danm thing away and download debian, but each to his own I guess;) 01:41
putter dduncan: ok, so we're in threads (blech;) plus readline land. hmm. do we have the equivalent of a ./configure --without-readline option? 01:44
dduncan I don't exactly know 01:45
I know that some programs do
and even the file in question ...
has an ifdef regarding readline
putter that would be the fastest way to get you unstuck (well, no, the fastest would be to comment out Readline.setCatchSignals and the next line in Pugs/Shell.hs. ;)
dduncan Shell that is
01:45 Limbic_Region joined
dduncan I just reverted Shell.hs, and that unstuck me 01:46
make is now successfully done
01:47 luqui joined
putter yeah. we dont (--wo-rl). Makefile.PL just checks has_ghc_package('readline') and based on that -DPUGS_HAVE_READLINE. though easy enough to add somthing if we go down that path. let's see... 01:47
dduncan but Shell.hs and/or some pugs config file will have to be updated to account for Tiger having a readline, but that is too old or too wrong 01:48
01:48 xinming joined
dduncan hm, there is something else I'll have to fix on my system too ... 01:49
the pugs tests are running about 3X slow again ... 01:50
luqui dduncan, when you start pugs shell, does it start instantly, or does it spend some time loading the prelude 01:51
dduncan just a minute
that would be the problem 01:52
oddly enough, lately running the pugs shell also causes a liteny of beeps
anyway, the prelude thing is easy enough to fix 01:53
putter dduncan: do you have some way of checking the version of your libreadline? locate libreadline and see if it's called libreadline.N.N.so? 01:56
dduncan one minute ...
fyi, I did say what the top line of my readline.h was
putter hypothesizes its 4.0 <= rev < 4.2 ... ;)
? 01:57
dduncan in usr/lib it says libreadline.dylib without a version ... looking further 01:58
that file is a softlink pointing to libedit.2.dylib 02:01
implying it may not be a real readline
putter actually, if you found readline.h, it should have a RL_READLINE_VERSION in it..
dduncan it doesn't 02:02
it says /*$NetBSD: readline.h,v 1.11 2004/01/17 17:57:40 christos Exp $*/
at the top
anyway, I have to go for a bit
pasteling "dduncan" at 24.108.164.7 pasted "my whole /usr/include/readline/readline.h" (190 lines, 6.5K) at sial.org/pbot/15652 02:04
dduncan maybe that will help you 02:05
I have to go to dinner now
putter mmm, food. ;) cheers &
02:07 drbean joined 02:19 xinming1983 joined
putter dduncan: I assume you are _not_ using darwinports? (that would raise some, apparently distinct, issues) 02:21
dduncan: did you compile ghc yourself, or use the binary from www.haskell.org/ghc/download_ghc_641.html ? 02:25
the latter explicitly requires libreadline.5.0.dylib. as for the former... 02:27
02:28 sub_chick joined
putter it's currently looking like tiger uses a netbsd readline.h (thanks to your paste); that netbsd maintains its own readline.h, and is not very systematic about it; tiger's is old, pre-dating the addition of rl_catch_signals ( archive.netbsd.se/?ml=netbsd-curren...p;m=868285 ); 02:30
ghc uses a simple "is it readline 4" ifdef (libraries/readline/System/Console/Readline.hsc); soo... 02:32
my working hypothesis is ghc is using a test which works fine if gnu readline has been directly installed, but not in the face of the complexity on tiger. pugs simply uses ghc's conclusion (Makefile.PL's if(has_ghc_package('readline'))), so we lose. 02:36
the core problem for pugs is we don't have a way to ask if ghc has a _working_ readline. 02:38
02:43 xinming1983 joined
putter so it's in the log, the key line from the paste is /*$NetBSD: readline.h,v 1.11 2004/01/17 17:57:40 christos Exp $*/ 02:44
02:44 BillN1VUX joined
dduncan putter, 02:44
I'm using the precompiled ghc from GHC-6.4.1.pkg.zip, linked to on the page you mentioned 02:45
I did not explicitly install libreadline first, though
I think I just assumed it was working when GHC executed and pugs otherwise worked, up to today 02:46
so now I just have to figure out where to get readline ... 02:47
putter right. ok, I propose: add a PUGS_DONT_USE_READLINE environment variable check to Makefile.PL, and mention it in a tiger-specific INSTALL section. the alternative is just to require tiger folks to install readline, and fail if they havent. 02:48
dduncan fyi, I was going to install the newer readline
as the way to fix the problem
assuming that doing so doesn't break something in OS X that depends on broken behaviour of the bundled one 02:49
putter right. :) I guess I've moved on to the "fix the install problem dduncan turned up". ;)
how much of a barrier to entry is installing readline for the "average user who wants to run pugs"? 02:50
dduncan well, the options are either to add that switch, or make the newer readline a hard dependency of running pugs
02:51 Shillo joined
putter right. what do you think? 02:51
dduncan hard to say
right now, I'm going to try and install readline, and see how difficult or not that is
I would probably need to install it anyway, since some other things like PostgreSQL would require it
putter ok, I'll add the flag. its localized enough cruft. the mere fact you are worrying about it suggests its a nontrivial item. 02:52
dduncan I know that readline is used by many things
but to be honest, from a user's point of view ...
I would prefer to install as little as possible in order to run pugs 02:53
Shillo dduncan: Plus, readline is strict-GPL. Kinda problem if you want something to be LGPL 02:54
... or artistic.
dduncan would it be simpler for pugs to default to not using readline, and only use it if an environment variable is set?
that way, saavy users do a bit of extra work to gain extra functionality that most people don't need 02:55
putter on the other hand, I could just read more carefully. Makefile.PL is already compiling a little test script to see if readline really works. I'll just add the signal commands to the test script. I'd re-underline the "note to self" which says "avoid tunnel vision when debugging", but that paragraph is already worn thru with underlines. sigh.
dduncan and non-saavy users can do as little as possible 02:56
but if as you say there is a way to test this in Makefile.PL, then great, do that
and in fact I *won't* go to the trouble to upgrade readline
I assume your test will say that if readline is broken, then it will be as if it didn't exist at all 02:57
I will wait for your commit ...
putter dduncan: thanks. I was just about to (belatedly) suggest that :) edit done, testing... 03:06
03:12 vytautas left 03:17 lypanov left
putter drat! I forgot to end the "was taking far to long to write" log entry with dduncan++. 03:18
Belatedly, dduncan++. Very.
svnbot6 r8826 | putter++ | Makefile.PL: The embedded haskell test program, which checks whether readline really works, now includes calls with the newly required (r8825) setCatchSignals and setCatchSigwinch calls.
r8826 | putter++ | The old test was passing on Mac OS X 10.4 Tiger, which doesn't support the new calls.
r8826 | putter++ | A result of all these changes is the average Tiger user, who hasn't installed readline herself (installation is suggested by the ghc download page), will no longer get the readline version of pugs.
putter dduncan: patch is in. if you could check whether it properly fails, that would be great. Makefile.PL should run as usual, and the resulting makefile should _not_ have -DPUGS_HAVE_READLINE in it. 03:20
dduncan I am downloading it now, but first I just changed my Shell.hs to the newest version
putter ah, good idea 03:21
dduncan fyi, rerunning the Makefile.PL does not spit out the "not installed" message 03:22
putter thought mac os x was a linux. what's it doing with a netbsd readline? does it use the netbst signal system? curious...
dduncan Mac OS X is derived from BSD, not Linux 03:23
its in the BSD family
by way of NextStep + OpenStep
putter "not installed" message? is that something which was there previously?
ah.
dduncan what I mean is, it doesn't say disabled 03:24
oh wait, yes it does
putter oh. that's not good. oh, it does. whew.
dduncan for some reason I expected it to come out before all the Writing makefile for ext/foo
but it came out after
putter it looks like Makefile.PL collects the warnings in a @warn list. 03:25
dduncan I've got the simplest possible pugs setup ... all optional extensions are turned off
no linking to haskell extensions, no readline, no perl 5, no parrot 03:26
currently on 'make'
putter well, you probably had readline before. now you dont. :/ 03:27
does the Makefile contain -DPUGS_HAVE_READLINE ?
dduncan I only switched to Tiger 7 days ago, so I'm used to not having it
putter ah
dduncan looking
putter nice to know NextStep wasn't a complete deadend. nice system. 03:28
dduncan the text 'readline' is not in the makefile at all
putter yay
dduncan yes, Mac OS X has lots of NextStep goodness
Shillo G'night, all! 03:29
putter good night &
03:29 revdiablo joined
dduncan the way that all native applications and frameworks are bundles of files, folders etc 03:29
the whole 'Cocoa' development api
which is the whole object-oriented api that NextStep started
in Objective-C
putter ah right 03:30
dduncan of course, what pugs and ghc is working with is beneath that, using the BSD api of mac os x, which is sideways of Cocoa
make has successfully compiled Pugs.Shell 03:31
putter woot.
dduncan thank you for your help
putter oh, thanks for finding the bug!
and debugging the bug! (that sounds odd) 03:32
if you debug a bug, what have you got?
dduncan: should we keep all the cruft in INSTALL re "until 6.4.1 comes out", or simply say, like x64_64, mac requires 6.4.1 ? 03:33
dduncan I'm using 6.4.1 03:34
putter s/mac/gcc4.0 and mac/
dduncan and have been since september
apparently 6.4.1 is the first version that works properly with Tiger
6.2.2 and 6.4.0 work with Panther 03:35
which is where I started
putter right. sorry - when working the last problem, I noticed INSTALL was still crufty, providing "until 6.4.1 comes out" workarounds. which should now no longer be needed. I'm wondering if I can just rip it out, whether you think it still has some value, and should say "6.4.1 recommended. But if you don't, ...<existing cruft>" 03:36
dduncan 6.4.1 has been out for 4 full months now ... it should be a hard requirement for pugs, such that the Makefile.PL refuses to proceed without it 03:37
we ditched 6.2.x when keeping that was holding us back, we can do the same with 6.4.0 03:38
putter ok, editing... 03:39
oh, wait. a hard, universal, requirement? hmm. 03:41
putter wonders if that's an audreyt call... eh, I'll do it as a separate patch.
03:44 BillN1VUX left
svnbot6 r8827 | putter++ | INSTALL: Simplified "until GHC 6.4.1 comes out" caveats for gcc 4.0. Now that 6.4.1 has been out for a while. 03:44
dduncan well, what good reason would someone install 6.4.0 and not 6.4.1? 03:45
unlike perl 5, ghc is usually not bundled with a system
people have to install it anyway
obra often people install packaged software
dduncan yes
and ghc 6.4.1 has been in package form for many months 03:46
obra And, well, packages aren't 100% up to date.
for every platform?
Don't be too cavalier with forcing people to latest versions.
dduncan I would think that any platform having a 6.4.0 package also has a 6.4.1 package
obra I'd check that assumption before possibly freezing people out
dduncan we already froze 6.2.x out 03:47
obra Bring it up on the mailinglist and see if it'll screw people over?
dduncan okay 03:48
obra Cool
putter ok, I'll sit on the Makefile.PL patch. Both it and INSTALL would be simplified by making 6.4.1 a hard dependency. 03:52
audreyt: thoughts?
dduncan okay, the p6c message is away 03:53
03:53 SamB joined
putter dduncan++ 03:53
Thanks, that was a fun exercise. :)
dduncan putter, if you want, you can send a reply to p6c saying you already have a patch ready and are waiting for the good word 03:54
03:54 Cryptic_K joined
dduncan putter, is 8827 just documentation/comment changes? 03:55
putter na, it's pretty trivial. besides, separating policy from mechanism is always a good call. this is a policy question.
yes
I removed the notes on how to jump through hoops until 6.4.1 is available. 03:56
s/always/usually/ 03:59
putter runs a smoke to see the ext/ fixes... 04:08
dduncan about 5 distros were affected by what I did yesterday, there are a half dozen or so left to fix
putter yay :) 04:09
putter seriously underestimated how long this release process would take. 04:10
04:10 luqui joined
dduncan this is certainly the longest releast yet ... about 4 months 04:10
luqui dduncan, why should ghc-6.4.1 be minimum? 04:11
is there a feature we're using
which isn't in 6.4.0?
dduncan it simplifies things for people using GCC 4
6.4.0 doesn't work with gcc 4
putter audreyt: once the release is out the door, it might be fun to have another conversation/interview, regards what the next steps are, what exactly they involve, and so on.
luqui dduncan, so?
people who aren't using gcc 4 can still use 6.4.0
dduncan so people with that will have to explicitly say to use a different version when compiling, rather than the default action working
you misunderstand 04:12
ghc 6.4.1 works with both gcc 3.3 and 4.0
ghc 6.4.0 works with only gcc 3.3 and not 4.0
luqui yeah, I understand that
putter luqui: the set of things which require 6.4.1, and thus have to be documented/tested separately, are gcc4.0 and thus Tiger, x86_64, embedded parrot, and... I think that's it. 04:13
dduncan if ghc 6.4.1 is required, then it doesn't matter what gcc people have
luqui so the minimum requirements seem to be: gcc 4.0 and ghc 6.4.1 _or_ gcc 3.3 and ghc 6.4.0
that's logic that can easily be coded in a configure script if necessary
I don't think *requiring* the installation of ghc 6.4.1 is making anyone's life easier
putter Oh, maybe not embedded parrot. I'm confused. Though my embedded parrot is also not working. But that's a separate issue ;) 04:14
dduncan the current minimum requirements are (any gcc and ghc 6.4.1) or (gcc 3.3 and below and ghc 6.4.0)
luqui dduncan, and what's wrong with that?
dduncan requiring ghc 6.4.1 means not requiring gcc 3.3
its a trade off
luqui so?
the reason you bump the required version is not to save yourself a couple words in documentation 04:15
dduncan since more people have gcc pre-installed, you have a net result that fewer people have to change something for pugs to work
luqui if we find a feature that ghc 6.4.1 has that pugs needs, then we can bump the minimum
dduncan, but what about people who already have ghc 6.4.0
dduncan what systems have ghc pre-installed? 04:16
wouldn't people with it have done it themselves?
luqui just a sec
dduncan if they could do it once, doing it again is easy
if they've never done it, then doing it once is hard
but if they've never installed it manually, they probably don't have it 04:17
s/manually/explicitly/
pasteling Someone at 67.165.197.242 pasted "new ghc policy requires strictly more work" (13 lines, 449B) at sial.org/pbot/15654 04:20
luqui in that table, where I classify all users according to their status
I can't find a single case where the new policy is less work
putter lol
dduncan what if you're only installed version of gcc is 4.0? 04:21
luqui that's the second line of those tables
dduncan then you have to install 3.3, which is more work
luqui no, you install ghc 4.1
er, 6.4.1
putter luqui: in a production environment, I'd agree with you whole heartedly. In an experimental system like pugs, the question is whether the value of simplicity and clarity wins out. the lower complexity. having, at least until 6.5 comes out, only one version of ghc to care about. 04:23
luqui putter, well, like I said, if we find a feature of ghc 6.4.1 that we think we need 04:24
then by all means, increase
but atm, 6.4.0 works fine for anyone who has gcc 3.3
putter wonders if we're having a conversation about order-1 (ie, 1-3) or order 10 (ie, 3 to 30) users. ;)
and doesnt have 64bit. 04:25
luqui well, I have gcc 3.3 and ghc 6.4.0
so it's at least one user
:-)
dduncan I note that the recommended process for installing ghc is using a binary package, and doing so only takes a few minutes ... how much trouble is that?
or do you do yours from source?
luqui I build mine from source
dduncan okay then 04:26
how long does that take?
luqui a long long time
dduncan I hear quite awhile ... days
luqui (i think last time it took about six hours)
putter not really the point. the question luqui is asking is, what is purchased by requiring this effort of mine. if the answer is a smaller INSTALL file, then it probably isnt worth it.
dduncan why did you go that route rather than the binary?
luqui putter, exactly
dduncan, dunno. for fun? there's really no reason I couldn't install the binary. I'm just arguing my point. 04:27
pugs isn't itself using any features particular to ghc 6.4.1, so why not enable a 1% increase in user base (even though 1% of the user base is less than one person :-) 04:28
and haskell is a lazy language. so we should probably have a lazy upgrade policy, too :-)
(what a strong argument!) 04:29
dduncan are there any features that pugs would desire to use but isn't due to backwards compatability?
luqui I think 6.4.1 was a bugfix release
I'm not sure though
putter tries to remember where the debian package stats are. now curious as to how many people are using pugs. and to provide a baseline for comparison with downloads of the next version.
dduncan and none of those bugs actually cause us trouble?
luqui pugs works for me :-)
putter luqui: re (what a strong argument!), and a funny one too! In a -Ofun environment, that carries extra weight. ;) 04:30
dduncan: you and I need 6.4.1 (tiger, x86_64). luqui doesn't (everything else). a hard requirement of 6.4.1 is zero work for us. luqui is standing in for everyone on everything else, and saying "why are you doing this to us???". ;) 04:34
as for the bugs, it's quite possible luqui will see bugs that we don't. if that becomes an issue debugging pugs, then that will be an argument for a hard requirement. 04:35
s/will see/is seeing/
audreyt dduncan: did the readline shell segfault for your (or cause infinite repetitions, etc) ? 04:38
s/your/your environment/
dduncan: that is, before my sigwinch change
dduncan I'll check
putter did fix the build though
audreyt yup, but I wonder if it makes sense to restore readline for you
dduncan oh, and running pugs shell no longer makes all those beeping sounds 04:39
audreyt aha
dduncan like from trying to output an unprintable character
audreyt so you _did_ run into that bug
with the netbsd readline
putter so far, I don't think there has been a sign of it. though it may be hidden in some of the things which have been mentioned as issues, like compilation speed/space, or embedding problems. no one asks/says which version of ghc is being used, because we haven't yet noted a need to be sensitive to that.
which isnt to say we aren't currently being burned by it. which is the simplicity argument.
dduncan I did run into the beeping since a day ago, and now it is gone'
audreyt hm, since a day ago.
dduncan I updated to Tiger 7 days ago 04:40
I rarely use the shell
audreyt ok
dduncan the problem may be a bit older, but I only remember it since 1-2 days ago
04:40 drbean joined
audreyt nod, I hit it at every Tiger installation 04:40
it's not as bad as segfaulting on freebsd here
but still quite annoying 04:41
ok, I think disabling readline for Tiger by default is sane, then
(i.e. putter++'s current treatment) 04:42
dduncan and the way putter did it, by checking for the symbol, is best as it allows Tiger people who installed newer readline themselves to use it 04:43
audreyt hm, let me check in a simple test 04:44
dduncan I was about to smoke now, but I can wait for that
or I won't
putter not so much checking for the symbol, but embedded test hs script again using all the functionality the real Shell uses. the problem was, the Shell's usage expanded without the test's doing so (and the domain turned out to be complex enough for this to matter). 04:45
dduncan this smoke will see what affects that audreyt's pile-o-changes made since my last commit
and then I will look into getting the rest of the ext/ to work 04:46
putter is currently smoking 8825
dduncan I'm doing 8827
putter k
dduncan on my machine at normal speed, it takes 2 hours
question ... 04:47
putter OSes as complex domains --
audreyt ok, r8828 is in.
dduncan if one sets config.yml to have 2 concurrent processes ...
how many pugs processes should be running at once during the tests?
audreyt 2~4?
dduncan I only ever see 1
putter hmm. me too, now that you mention it.
svnbot6 r8828 | audreyt++ | * Readline emulation in Pugs.Shell for people without 04:48
r8828 | audreyt++ | GNU readline in their system.
dduncan I see 2 perl processes using zero cpu
but only 1 pugs process open
audreyt dduncan: can you try r8828 and shell?
dduncan so it isn't really using both of my cpus
do you want me to kill my smoke first?
audreyt dduncan: no, not really, ^Z will do
04:48 Amnesiac joined
putter oh, wait. I'm not setting config.yml. never mind. 04:49
dduncan well, my config.yml contains "smoke_concurrent: 2" and that value is reflected at 'make' time 04:50
but I still only see 1 pugs process when smoking
putter btw, hi audreyt. :) good dinner with the folks?
dduncan this has been the case for quite awhile now
Pugs.Shell did not die when being compiled 04:51
audreyt putter: yup, people are quite adaptable :)
dduncan on r8828
otoh, when compiling Pugs, it did die
putter yay :)
dduncan a whole bunch of undefined symbols "_SystemziConsoleziSimpleLineEditor_*" 04:52
audreyt, did that answer your question?
audreyt dduncan: yes. please hand-hack Pugs.cabal
dduncan note that I didn't rerun the Makefile.PL, only 'make'
audreyt or rather Pugs.cabal.in 04:53
before "QuickCheck -any"
in line 7
add a
readline -any,
to it, and "make pugs" again
dduncan okay, I have pugs.cabal.in open
audreyt line 7 should be "build-depends: ...", add "readline -any, " as one of the depends 04:55
putter notes there is already a readline -any at the end of the line
dduncan I added the readline -any
putter, not on my line 7
clkao i so want something to tell me all possible occurances of $o->bar calls where $o can be Foo
putter build-depends: QuickCheck -any, base -any, mtl -any, stm -any, haskell98 -any, network -any, template-haskell -any, unix -any, readline -any
audreyt putter: yes, and you are looking at .cabal not .in
putter duh. tnx 04:56
dduncan audreyt, did you mean run 'make' or 'make pugs'?
after that edit?
will try 'make pugs'
audreyt both should do 04:57
putter clkao: maybe next release? though that will require having type annotations on the ast, so maybe not.
clkao i mean in perl5
dduncan that didn't work; same link errors
clkao *sigh*
putter lol 04:58
clkao i was ---><---- this close to write some analyser using ppi
pasteling "dduncan" at 24.108.164.7 pasted "my current (edited) Pugs.cabal.in" (26 lines, 2.2K) at sial.org/pbot/15656
dduncan audreyt, in case that is helpful 04:59
clkao right now i am totally depending on svk's test coverage so get those calls in runtime and fix them
putter what's the nature of the problem? 05:00
audreyt clkao: welcome to the "dynamic even if not needed, static quite impossible" language world
clkao the nature of the problem is i had some crappy interface. now lots of codebase is using them. and i am silly enough to do some refactoring when i am on holiday
putter lol :) 05:01
audreyt dduncan: so, pugs gets built, and you were able to use the shell? does the common line editing keys (home, end, cursors) works for you?
dduncan no I wasn't
after editing the file you mentioned, I get the same link errors as before on attempting to make
clkao audreyt: we need to write this on ꘄčÆ
putter: so in short, i suck 05:02
audreyt dduncan: ok, bad idea then, I'll revert my change.
thanks for the test :)
dduncan okay
maybe I did something wrong, but I thought it was what you told me to do 05:03
05:04 drbean joined
putter clkao: hmm, if there is a test suite, one could try the (ideally automated) edit one/some instances of ->bar, run test and see if it fails, repeat for all ->bar. assuming $o not Foo has a detectable signature (ie, a test fail) 05:04
audreyt dduncan: you didn't do anything wrong, I think that's as expected
no worries :) 05:05
clkao ya. that's more or less what i am doing. but i am depending on the test coverage
how would perl6 make my life easier in this case?
i assume it's easy for some tool to tell me all $o that is Foo right?
audreyt clkao: type inferencing, mostly 05:06
putter could filter the set of interest $o's by excluding ones whose type is known to be incompatible with Foo.
audreyt clkao: which is that "some tool", yes
clkao *nod*
svnbot6 r8829 | audreyt++ | * revert the failed attempt at emulating readline
r8829 | audreyt++ | using System.Console.SimpleLineEditor, as it didn't
r8829 | audreyt++ | quite work for dduncan++ (and neither for me)
audreyt but it does require some annotations on your part.
e.g. declaring classes with "class" and add types to member variables and function signatures etc. 05:07
the more annotations you put, the more information the inferencer gets
clkao yep
putter could also do semi-automated "aspect oriented programming", ie, massage questionable instances of $o into $o.does(Foo) ?? {log_oops;$o->before} !! $o->after 05:08
audreyt lexwrap can already do that for perl5 05:09
clkao i think the point is to get those in compile time 05:10
putter oh, I was afraid of that. meant $o.does(Foo) ?? {log_oops;$o->bar0} !! $o->bar1
or shouldnt I be looking at Hook::LexWrap?
dduncan the reverted builds 05:12
audreyt putter: just installing a Foo::bar prehook that does log_oops would do, no? 05:13
putter oh, I get it. too many different oo languages swirling in my brain.
clkao: so you could annotate / lexwrap all the ->bar's in your program, have the log the caller file/line, run you tests, perhaps if you have a captive user base, have them run it to, and then you'll have a plausible list of $o's to change. 05:16
clkao ya. but i also want to find those not covered by tests 05:17
putter thus the idea of giving users the instrumented version, assuming that's plausible. then you get the log from their real use. 05:18
hey, have smoke 05:21
audreyt woot
audreyt starts triaging based on putter's smoke
putter m19s28.vlinux.de/iblech/stuff/pugs-...3593d.html 05:22
and dduncan is running an r8828 osx. 05:23
clkao: re holiday/suck, as long as it's fun? if not, then -Ofun ;) 05:27
putter ponders the meaning of calling ones' self with alternate arguments... 05:28
dduncan actually, I'm running r8829 now 05:30
but I expect the test failures to be about the same
as far as audreyt's triaging goes, you should focus on the t/ items 05:31
audreyt okay, doing so 05:32
dduncan because ... if they are correct, then any ext/ failures are probably due to bad ext/ code and not bad pugs code, which makes things easier for me
audreyt nod. 05:33
dduncan that said, I think Pod-Event-Parser was working before your changes
but I will confirm
audreyt okay.
svnbot6 r8830 | audreyt++ | * junction/chained_operators.t: unbreak the test syntax. 05:34
audreyt I'll be offline a bit to focus on triaging
will be back in 0.5~1hrs
05:41 xinming joined 05:56 drbean_ joined 06:05 drbean_ joined
dduncan maybe I just need to configure my terminal, but the current Pugs doesn't accept the delete key 06:07
hitting it just causes junk to display
when in shell
mind you, the delete key works in other terminal apps 06:08
svnbot6 r8831 | audreyt++ | * TODOing variously dying oo tests.
dduncan so, Config::Tiny fails to parse, whereas it did parse yesterday 06:09
it appears that removing the line "$line .= chomp;" allows it to parse
currently checking that in isolation ...
06:10 xinming joined
dduncan ?eval my $x = "fu"; $x .= chomp; 06:12
06:12 evalbot_8807 is now known as evalbot_8831
evalbot_8831 Error: unexpected "." expecting operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input 06:12
dduncan so that parsed yesterday, and now it doesn't
?eval my $x = "fu"; $x .= chomp(); 06:13
evalbot_8831 Error: unexpected "." expecting operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
dduncan ?eval my Str $x = "fu"; $x .= chomp; 06:14
evalbot_8831 Error: unexpected "." expecting operator, postfix conditional, postfix loop, postfix iteration, ";" or end of input
dduncan ?eval my $x = "fu"; $x = $x.chomp;
evalbot_8831 \"fu"
dduncan the core test t/pugsrun/03-dash-p.t is also dying, and uses .= chomp 06:19
oh wait, that's in the pod of the file 06:21
so its not the cause of dash-p's death 06:22
putter dduncan: I noticed chomp.t is being skipped.
dduncan oh? 06:23
putter don't know why. looking at log...
dduncan oddly enough, chomp.t itself isn't being skipped, but for its last 2 tests 06:27
chomp.t hasn't been updated in 3 months 06:29
various evidence points to the Parser changes made yesterday as the culprit, assuming that $foo .= chomp is valid perl 6 06:30
putter someplace between r8818 and r8825, chomp.t became a parsefail
unexpected "." expecting operator, postfix conditional, postfix loop, postfix iteration, ";" or "}" at t/builtins/strings/chomp.t line 24, column 10 06:31
dduncan it was at r8824-5
my posted smoke of 8823 shows it passing 06:32
as well as various chomp using things passing
putter t/syntax/inplace.t is like that too. and there are a couple of other candidates. 06:33
dduncan its r8824, the big mass of triaging, that broke it
putter ah, ok
dduncan that change included several parser file changes 06:34
putter ok. maybe wait until audreyt surfaces. may already be fixed. ;)
dduncan mayhaps ... but who better to fix something than the one who broke it? 06:35
um, ...
06:35 xinming1983 joined
dduncan you do agree that "$x .= chomp;" is correct, yes? 06:35
and r8824 wasn't just exposing bad code? 06:36
putter re agree, so chomp.t suggests
dduncan well, I probably won't be testing any more today, waiting for that fix to come in first
putter shift change ;) 06:37
dduncan also, unless the issue is transient, which I'm testing now, there are issues using the pugs shell now 06:38
putter oh?
dduncan eg, backspace/delete key doesn't work
putter it's *nix. bs/del are _supposed_ to not work. ;)
sigh. 06:39
dduncan so what key do you hit to delete something you just typed into the pugs shell?
those keys work fine in the ordinary shell, and other cmdline apps
anyway, I did a make realclean and am trying building again, in case that experiment with editing cabal left crumbs around 06:40
putter since you're not seeing readline any more, if (haven't checked) pugs is just reading in line-at-a-time mode, then in editing you're just seeing some flavor of your terminal configuration, no?
dduncan well, the delete key worked in line-at-atime mode before 06:41
under Panther
putter I suspect you were getting readline back then. the downside of the pugs readline changes, is now you dont. 06:43
dduncan you think I was getting readline in Panther?
putter why not?
dduncan somehow I thought it was added for Tiger
in any event, that stuff we went through today ... 06:44
meppl guten morgen
dduncan about readline
I was under the impression I had never been using it before, under Panther
putter tries to remember how to say good morning in some language other than english. fails. sign it's time to say good night.
dduncan and the delete key is so basic, that even a non-readline input replacement should support it 06:45
meppl good night putter
putter no, the chain goes like this: pugs was using readline whenever ghc saw it, and a little test script passed. but there were issues with threads. those were fixed by using more of the readline api, which happened to not be supported by tiger. causing build failure. so the test was extended to check for the new features. which 06:47
dduncan but I thought readline was just for providing command histories and such
putter which fixed the build, but still means that the pugs readline implementation has changed in a way incompatible with current and past osx. so now osx isnt getting readline any more, unless you install libreadline yourself. 06:48
dduncan very basic things like typing characters and erasing characters should be in any input method
putter you would think so. but that's not the *nix tradition. ;) 06:49
pugs is in fact falling back to a simple getLine. so its a ghc + terminal configuration problem. 06:50
dduncan so what then, did unix people never erase characters?
putter meppl: :)
dduncan if what you say is true, then why can my same terminal still work properly with those control characters? 06:51
putter so it's a hairy mess which often doesn't work.
06:51 iblechbot joined
putter I would say ioclt, but part of the fun is each version of *nix has its own nuttiness. I (happily?) dont remember what bsd's flavor is. 06:53
perhaps I'm being too negative. perhaps posix has improved things. but this corner of the world, terminal control and configuration, and the inability to get del and backspace reliably working, has been a problem for more than a human generation. it gets a bit irritating after a while. smells like dead fish. ;) 06:54
06:57 nothingmuch joined, Alias_ joined
Alias_ ping anyone in the know? 06:57
putter a less obscure answer is that terminals have modes, the modes and the way they are set vary between versions of *nix, and often have obscure nonlocal behavior and dependencies, and thus are often not used to correct result.
Alias_ I just saw the announce of Haskell' 06:58
So, any other lambdacamels on the committee other than audrey?
putter looks over sholder. notes audreyt is off to focus on hacking, will be back eventually.
Alias_ Is Perl invading Haskell :)
putter I don't remember noticing any perl names other than audreyt. then again, i didnt really look. so ?? 06:59
rather than invasion, perhaps a venn-diagram community intersection? based in part on ghc's use of perl, and in part by audreyt sucking people into pugs? 07:00
Alias_ indeed 07:01
It's an unbelievably elegant merger
putter good description
Alias_ The more pure, hardcore, computationall strict language, built up from pretty much pure math, and the most flexible, scripty DWIM language
Meld the two close enough and you can pretty much do it all from military crypto to obfuscated oneliners 07:02
putter camel mates with ....
?
Alias_ a bigger brain?
We need a camel with a swolen head 07:03
putter oh my. that's quite an image
Alias_ Or pure math... which would be a camel in black leather in the Matric
Matrix
putter a while back there was the oneliner of "perl6 is the mechanism by which ML?Haskell? tries to take over the world". 07:06
Alias_ ML?
07:07 xinming joined
putter en.wikipedia.org/wiki/ML_programming_language 07:07
Alias_ The predecessor to Haskell or something 07:09
O! Caml!
putter yeah. yes :)
and sml. though caml seems to have the more active community. 07:10
Alias_ Well, it wins those programming competitions 07:11
putter wonders how hard it would be to have ml or hs on top of p6...
Alias_ And that's about all I know
putter caml.inria.fr/
:)
nice web page redesign. much simpler. was very cluttered. 07:12
07:15 ayrnieu joined
nothingmuch morning 07:17
putter g'morning 07:19
meppl good morning nothingmuch 07:20
putter Alias_: one nice use of ocaml: manju.cs.berkeley.edu/cil/ 07:25
Alias_ wow 07:26
But then C analysers have had a LONG time to evolve, and a LOT of money thrown at the problem
putter yeah, but it's only in the last very few years that open-source solutions have started to exist. Cil, OpenC++, gcc with tree dumps. sucks when everything takes multiple decades. 07:30
Alias_ true 07:31
putter it at least very vaguely looks plausible that hugs could be massaged into p6. cvs.haskell.org/cgi-bin/cvsweb.cgi/hugs98/src/
Alias_ I have a code generation/refactoring system for Perl, and when I was researching this about 4 years ago, the only BIG refactoring/analyser stuff I could find for C was half a million per seay
putter yeah
Alias_ seat
07:32 xinming1983 joined
Alias_ And god knows how much CPU needed 07:32
putter arpa et al kept funding individual researcher's efforts. while encouraging their commercialization. sometimes a good policy. but here it meant there was one early limited unsupported public version, and a almost always unsuccessful attempt to commercialize it. leaving nothing. there were at least 3? cases like that. :( 07:33
then again, it's a non-trivial exercise. especially C++. putter wants to see lots and lots of languages on top of p6. at this moment exploring haskell. But C++...? "run away! run away!" 07:36
Alias_ heh 07:38
putter hmm... I wonder how hard it would be to do big refactoring / analyser stuff in p6... half-million you say... ;)
Alias_ But oh the libraries we'd get
putter re "But oh the libraries we'd get", that's the "expendable-extra-character choosing to linger to gather more riches" line. projects head off down that pretty path... and their wreckage washes up years later. 07:40
Alias_ oh wait, C++ 07:41
We'd get buffer overflows and such inherited?
Surely not :)
putter s/riches/riches, while the tomb collapses/
07:41 Aankhen`` joined
putter its more that its really really complex. so for instance, M$ intentional progrming folks (pour code into database) did C++ because M$ does C++. several rewrites from scratch later, and n years, the project was eventually shot. 07:44
luqui putter, dost thou havst an example?
of said riches
putter lol :)
luqui er, said wreckage'
oh, you just gave one 07:45
yeah, C++ is a pretty damn complex language 07:46
Alias_ luqui: Pour example, native KDE integration
The riches!
Integration with games programming
luqui Alias_, why can't you use NCI for that? The mangling?
Alias_ Perl 6 as the language for Quake 5 modding
anwyas
Alias_ wanders off 07:47
07:47 Alias_ left
luqui Well, Lua saw the light for WoW 07:47
putter ha! Alias_ is sneaking off to gather riches which are rightfully ours!
WoW?
dduncan I'm going to investigate some of these terminal/readline/whatever issues 07:48
will try rebooting under previous my previous OS from last week, to see if it's normal again 07:49
07:49 dduncan left
putter fondly recalls working with dduncan, last seen heading hopefully down a dark and deep rathole 07:49
putter heads off for the accessible riches of sleep 07:50
'night all &
Aankhen`` G'night.
07:57 xinming1983 joined
meppl good night aankhen 08:02
08:03 dduncan joined
dduncan okay, following up with that test 08:03
I've confirmed that, ... 08:04
A: the current pugs compilation has the same shell problems under Mac OS 10.3.9 as it does under Mac OS 10.4.4
08:04 Cryptic_K joined
dduncan B: 10.3.x *never* had any version of readline installed ... 08:05
and yet the pugs shell worked properly there
a few days ago
likewise, I seem to recall but not sure it worked normally under 10.4.4 a few days ago
seeing the same broken current pugs behaviour under the terminals of both versions seems to rule out that a terminal difference or config is the problem 08:06
also, the fact that deletes et al worked properly in pugs a few weeks ago, when readline was never installed, says that readline isn't a prerequisite for such simple features 08:07
which means that the fact my pugs isn't linking in the older readline on my 10.4.4 system shouldn't cause a failure
so there would seem to be a problem in the recent pugs shell itself 08:08
and this should be fixable
... I will look into some file diffs further ...
08:11 elmex joined 08:22 xinming joined
dduncan comparing Shell.hs from revision 8162, it appears that r8820 may actually be the cause of my problems 08:22
I think it may be the new linebuffering commands that were added 08:23
integral oh, Makefile.PL has been changed recently :-/
dduncan they appear in the code unconditionally whether readline is in use or not; I will try moving them into the ifdef for readline
I don't think the Makefile.PL has anything to do with my problem
clkao hmm, jsperl5 is passing less. weird.
dduncan the problem I'm seeing right now isn't a make linking problem, but a runtime input/output buffering problem 08:24
integral ah, sorry, still thought we were back on that
dduncan I'll see what moving those lines does
no, the linking is fine
actually, revision 8825 has those commands in the ifdef 08:26
I think that autrijus' subsequent experiment and rollback in 8828 and 8829 put the commands in the wrong place 08:27
specifically, 8829 is at fault, I think
will test that theory ...
I think I fixed it ... committing 08:38
at least, the change works for me, and it is based on an older version
push done 08:42
as r8832 08:43
whomever's around, please update to r8832 and confirm that my Haskell editing didn't break something 08:44
svnbot6 r8832 | Darren_Duncan++ | r2186@darren-duncans-power-mac-g4: darrenduncan | 2006-01-29 00:41:00 -0800
r8832 | Darren_Duncan++ | src/Pugs/Shell.hs : updated the expressing setting initializeShell so it only invokes NoBuffering when we're using readline, rather than doing it unconditionally; this fixes a bug seemingly introduced in r8829 ... as a result, the pugs shell now handles the backspace/delete key correctly
dduncan I don't think it did, but this is like the second time ever that I edited haskell
I need two testers, one that has readline and one that doesn't 08:45
for myself, the delete key works again 08:46
okay, so its almost 1am here ... good night 08:48
09:02 bernhard joined 09:10 drbean joined 09:19 G2 joined 09:31 avar is now known as avar_packing 09:36 avar_packing is now known as avar 09:45 beppu joined 09:54 elmex_ joined 10:07 chris2 joined 10:09 r0nny joined 10:22 feng joined 10:23 kane_ joined 10:36 beppu joined 11:02 christo joined
nnunley www.iro.umontreal.ca/%7Eboucherd/ms...es-en.html 11:09
svnbot6 r8833 | audreyt++ | * TODOization of three more tests.
audreyt dduncan: "foo" and "return foo" have different types in haskell, so I removed an extra "return" 11:14
svnbot6 r8834 | audreyt++ | * Unbreak dduncan++'s editing for readline users.
audreyt dduncan: thanks for catching my misedit for nonreadline branch
11:29 drbean joined 11:37 christo left 11:51 elmex joined 13:12 putter joined
putter dduncan++ 13:12
putter shall endeavor to be less pessimistic in future. 13:13
audreyt: hi. you saw dduncan's note re r8824 and chomp.t et al? 13:14
anyone: re haskell on p6, ghc's ext-core seems a plausible way to go www.haskell.org/ghc/docs/latest/htm...-core.html 13:21
13:34 verve_ joined 13:42 drbean joined 14:00 iblechbot joined
svnbot6 r8835 | audreyt++ | * TODOize and corrected the rest of main test suite. 14:09
audreyt greetings 14:10
putter: GHC's Core is less pleasant than (say) YHC or JHC's. 14:11
I think YHC is most promising
but then, now that I grokked the IL in the not-not-ml paper, it seems plausible as well
though there's no production-grade compiler from anything to that IL yet 14:12
14:20 marmic joined
svnbot6 r8836 | audreyt++ | * "chop" returns chopped string instead of modifying it 14:24
r8836 | audreyt++ | in-place, so remove the "is rw" from its signature.
r8837 | audreyt++ | * -n and -p now always surrounds the remaining flags with 14:28
r8837 | audreyt++ | while () {...} blocks, instead of only the -e parts.
audreyt ok, I _think_ that's all of the main test suite
audreyt starts another round of smokes
clkao whoot
audreyt the anniversary release looks hopeful
though I spent far too much time grokking the not-not-ml paper
very enlightening, though
bestian++ # helpful on the other side of curry-howard bridge 14:29
svnbot6 r8838 | audreyt++ | * "$foo .method" now parses again; "$foo. method" now parsefails again.
14:39 xinming joined
putter audreyt: would you like any of the other backend smokes run in parallel with yours? 14:46
audreyt putter: sure, a pil2js run would be nice
putter will do
audreyt putter: and no, what about chomp.t?
putter colabti.de/irclogger/irclogger_log/...l=310#l493 14:47
audreyt I accidentally fixed that in r8838 14:49
xinming audreyt: did you have a nice day yesterday? :-) 14:50
audreyt xinming: yeah, got plenty of red envelope too
xinming I spent a lonely day yesterday. :-(
audreyt learned a lot about cosmetics and clothings etc from my aunt
xinming: ow :/
xinming hmm, is preparing to go KunMing. I'll go in 2 weeks maybe. 14:51
audreyt how's kunming nowadays? I've never been there, only read about it in books
clkao audreyt: i did a pil2js run already 14:52
but it's a bit eaerlier
14:52 chris2 joined
xinming audreyt: Weather in KunMing is the most important case for people choose there. :-P It's all spring, no summer, no winter, and even no autumn. :-) 14:52
audreyt: You only need 2 suits to live there. :-P 14:53
audreyt wow, constant weather
sounds lovely
wolverian sounds boring ;) 14:54
putter thinks it sounds boring (obligatory US New England opinion')
wolverian (I'm just trying to stand up for the weather-from-hell we have here.)
Odin- lives where there's no weather.
putter where?
Odin- Just samples of weather. :)
audreyt weatherlets 14:55
xinming I grow in KunMing, and then back to WenZhou. It's a pain to live here at first... as I don't know when I need to change the clothes,
putter: No, It won't boring, You'll love there if you live there for a long time. :-) 14:56
s/won't/won't be/
putter :)
re haskell on js. yhs has the advantage of a simple vm, and it looks like someone is meeting with some success doing a .net for it. disadds include looks like a one-person project, and is still a bit toy-ish. jhc has a larger community, but is also far more ambitious. it would be nice to have acceess to its analysis phases in p6, 15:00
and it seems to be working on some snippets, but scaling issues make self-compilation look somewhat distant. ghc... is there some disadd to ghc core? other than the complexity required to run it? and its perhaps? lack of a user community? certainly ghc is the most capable of the compilers. whether that gets reflected in the core code is a separate risk point. 15:03
15:05 vel joined
audreyt putter: YHC is NHC's successor, so it has the sane NHC base 15:07
putter: GHC is the most capable, but GHC's core is not as easy to work with.
putter: certainly a JVM and a CLR backend existed for GHC, and we can work fron there.
but YHC's VM instructions does look very accessible.
but I need to sleep now :)
audreyt pressed "shift"
journal up... see you tomorrow. 15:08
putter ;)
good night audreyt &
Juerd audreyt: Sleep well
15:09 GeJ joined
putter www-users.cs.york.ac.uk/~ndm/yhc/ repetae.net/john/computer/jhc/ www-users.cs.york.ac.uk/~ndm/yhc/do...tions.html 15:10
integral oooh, dduncan++ # rewriting File::Spec 15:14
putter a build issue: I've been noticing that make builds a precompiled js prelude, but then make smoke-js builds one again. so something isn't quite right. 15:17
wolverian hmm. dduncan++ # nice 15:18
I'll rewrite Path::Class on that as soon as possible :)
integral wolverian++ 15:19
15:22 xinming joined 16:04 apple-gunkies joined 16:08 nnunley joined 16:24 nothingmuch joined 16:25 xinming joined 16:35 Amnesiac joined 16:53 justatheory joined
xinming anyone here use debian testing? 17:30
Welcome to Pugs -- Perl6 User's Golfing System 17:31
Type :h for help.
Loading Prelude... done.
Segmentation fault
I compiled this on both laptop and my pc. with the same result. and I've tried to make clean. :-/
ingy hola 17:39
seen audreyt
jabbot ingy: audreyt was seen 2 hours 31 minutes 1 seconds ago
obra ingy!
how's things?
17:48 weinig joined
clkao ingy, your mail is fscked. jo sent a mail to you 17:52
18:03 dduncan joined 18:05 elmex joined 18:12 pjcj joined
putter r8838 smokes for both freebsd and linux are up. core isn't quite green yet, but getting there. 18:19
18:26 pjcj_ joined 18:34 elmex joined
xinming hmm, anyone here got my problem? 18:40
failed to run pugs
18:55 pjcj joined
putter failed to run pugs? 18:57
gaal yo 19:14
pugs is still Rat-happy. Should this be fixed for the release? 19:15
theorbtwo Rat-happy? 19:16
integral bleh, my pugs really doesn't like t/pugsrun. /me kicks his box
gaal ?eval 10 ** 3 / 2 ** 10
no evalbot? :-( 19:17
19:17 evalbot_8831 is now known as evalbot_8838
evalbot_8838 125/128 19:17
integral it's just slow when it has to upgrade itself ;)
gaal ?eval "I wanted a decimal: " ~ 10 ** 3 / 2 ** 10
integral: I don't think in updates lazily ;-) 19:18
evalbot_8838 "I wanted a decimal: 0.9765625"
gaal feather must be loaded
theorbtwo What's wrong with that being a rat?
?eval 3/6
integral oh, damn I forgot to nice my make
evalbot_8838 1/2 19:19
19:19 phydoxx joined
theorbtwo ?eval 6/3 19:19
evalbot_8838 2/1
gaal it's not morally reprehensible or anything, but it's surprising 19:20
theorbtwo It is?
gaal sure, do the same code in Perl 5
theorbtwo Right. This is better. 19:21
gaal theorbtwo: I'm not sure it is; when the fractional parts get big it loses redability: 19:22
?eval 1000 ** 3 / 2 ** 1000
theorbtwo gaal: Only if you do .perl.
evalbot_8838 1953125/20927902484106783612273926739453160362527437728623703270385749772858418967283908642445280836244059729054583455420959898929436431361178008664032378075583153913934702685203576143400536338012443636480379262017668896452308479037888217888995203019681763505021868120481527671211777014946532005541417320448
theorbtwo ...and the two sides are rel prime.
gaal right. perhaps it should be "fixed" in prettyVal only. 19:23
the fraction quoted above gives very little cue about its size 19:24
theorbtwo If you don't like it, feel free to ~.
Odin- That's a problem with rational representation in general...
theorbtwo ?eval ~(1000 ** 3 / 2 ** 1000) 19:25
gaal "indeed" xx 2
evalbot_8838 "0"
theorbtwo Um... now /that/ is a bug.
Odin- Mmmm. Yes.
Quite certainly.
gaal should it be a bigsmalldecimal? 19:26
?eval [+] (~(1000 ** 3 / 2 ** 1000) xx 1000) 19:28
evalbot_8838 0.0
gaal oops, silly test
see, I don't know how to "shake off" Ratitude except by stringification 19:29
theorbtwo ?eval 1.0 * 1000 ** 3 / 2 ** 1000
gaal but you know what, maybe I shoudln
't care.
evalbot_8838 1953125/20927902484106783612273926739453160362527437728623703270385749772858418967283908642445280836244059729054583455420959898929436431361178008664032378075583153913934702685203576143400536338012443636480379262017668896452308479037888217888995203019681763505021868120481527671211777014946532005541417320448
gaal theorbtwo: not so easy :) adding integers doesn't help either. 19:30
theorbtwo ?eval 1.0 * (1000 ** 3 / 2 ** 1000)
evalbot_8838 1953125/20927902484106783612273926739453160362527437728623703270385749772858418967283908642445280836244059729054583455420959898929436431361178008664032378075583153913934702685203576143400536338012443636480379262017668896452308479037888217888995203019681763505021868120481527671211777014946532005541417320448
xinming putter: I'll get seg fault when I start pugs 19:33
theorbtwo Hm.
xinming Welcome to Pugs -- Perl6 User's Golfing System
Type :h for help.
Loading Prelude... done.
Segmentation fault
xinming@HomePc:~$
theorbtwo Possibly a bit of an issue. 19:34
gaal xinming: what OS are you on?
xinming: how much RAM do you have?
xinming debian 19:35
512M
I am in debian testing
gaal and, are you on the last r of pugs? audreyt fixed a segfault lately
I'd try playing around with heap size: 19:36
./pugs +RTS -M300 (I think it goes)
xinming I use the latest version
gaal: ?? Not works.
gaal hmmm, i thought there was an infectuous way of telling any ghc-compiled program how much heap to give, but I may be imagining things 19:37
xinming gaal: by the way, Is ghc 6.4 still supported for now? 19:38
gaal yes
xinming: try seeing what strace says? 19:39
strace -o /tmp/pugs.trace ./pugs ; tail -50 /tmp/pugs.trace
19:40 nicbrown joined
lisppaste3 xinming pasted "last 50 line post" at paste.lisp.org/display/16184 19:43
gaal hmm, nothing obvious 19:46
(except that haskell io is way inefficient :-)
actually, i wonder, why is it trying to open /usr/share/perl/5.8/strict.pm ? :) 19:47
xinming: oh you probably have embedded perl5, right?
xinming yes 19:48
gaal well, sorry, I can't think of anything. :(
xinming It's ok, I'll ask audreyt about this
putter xinming: have you tried it without embedded perl? 20:05
20:18 Limbic_Region joined 20:28 amv_ joined
xinming putter: I compiled it without the perl5 embed, There will be segfaults too. :-/ 20:43
putter :( 20:51
clean build? no bits left over from a previous one? 20:52
xinming putter: yes. I've even rm the dir and co again. :'( 20:59
putter smoke of pil2js r8838 is up. 21:22
xinming: what os, cpu, etc?
version of ghc?
21:27 lypanov joined
xinming I use ghc 6.4, and cpu is barton 2500+ from AMD, and using debian testing. 21:31
gcc version is `gcc version 4.0.3 20051201 (prerelease) (Debian 4.0.2-5)`
22:24 drbean joined
Juerd gaal: What did you mean by "loaded" in "feather must be loaded"? 22:34
putter xinming: ah, gcc 4.0. that requires 6.4.1. there's a note in INSTALL.
xinming putter: Ok, I think I have compile ghc myself. :-/ I'll try that, thanks 22:35
putter xinming: www.haskell.org/ghc/download_ghc_641.html has a debian binaries section. 22:48
(I googled "debian ghc 6.4.1") 22:49
xinming putter: hmm, I use testing, If I upgrade to 6.4.1, I have to upgrade many packages which are in unstable. :-/ 23:00
putter: what Os you use by the way?
freebsd? 23:01
23:10 feng123 joined 23:11 grayson joined 23:13 beppu joined
gaal Juerd: evalbot was responding slowly. somebody forgot to nice their build but they reniced it when they noticed. 23:13
?eval "am I slow now?" 23:14
evalbot_8838 "am I slow now?"
gaal I think now it's okay.
Juerd gaal: Ah, okay. Thanks for reminding this somebody to nice their processes 23:18
It's amazing how many people use unixlike systems at home, while having little or no experience with unix systems shared with other people :) 23:19
I've been considering upgrading feather, by the way 23:20
mugwump hardware or OS?
Juerd Hardware. I do OS upgrades often.
Odin- What's feather running, OSwise? 23:21
Juerd Odin-: Debian GNU/Linux
Odin- Which version?
Juerd Odin-: Sid
rafl would appreciate that. 23:22
Juerd I'd like a dual cpu system, possible amd64 23:23
s/ble/bly/
gaal RAM is probably the first priority, since if two people compile at the same time the machine risks thrashing
zZZ & 23:24
Juerd gaal: Most of the time, and that's more than 95% of the day, half of the memory is available. 23:25
gaal: I think making builds faster (and thus making memory available again faster) is more economic
Oh, just to make things clear: don't expect such an upgrade soon :) 23:28
gaal Juerd: if just *over* half is available then, fine, but otherwise consider that the danger band is longer than 5% of the time 23:29
gaal has no expectations, just cheery hopes regarding everything Perl 6
now where was that bed
Juerd Hm, feather has only 1 GB of memory. I really thought it had 2 23:31
Let's see what this board is capable of
"Support for up to 2 GB system memory" 23:34
Not much, but more than current
Bank: "DIMM0" 23:38
Size: 512 MB
Bank: "DIMM1"
Size: 512 MB
That's unfortunate, though 23:39

Powered by Cro and the RakuĀ® Programming Language.

Please report any issues / comments / feature requests as an issue on App::Raku::Log.

Thank you!