r25895 | chromatic++ | trunk: : [Lua] Fixed a couple of memory leaks in LuaString PMC. diff: http://parrotvm.org/svn/parrot/revision/?rev=25895 r25896 | chromatic++ | trunk: : [lua] Fixed several memory leaks related to exceptions; removed unreachable : code after exceptions. diff: http://parrotvm.org/svn/parrot/revision/?rev=25896 % slightlyoff has joined #parrot r25897 | chromatic++ | trunk: : [Lua] Ensured that every place that creates a new Lua PMC sets the appropriate : mark/destroy flags. Some clone() calls didn't. diff: http://parrotvm.org/svn/parrot/revision/?rev=25897 % cognominal_ has left cognominal_!~cognomina@82.67.232.89 % cognominal_ has joined #parrot % cognominal_ has left cognominal_!~cognomina@82.67.232.89 % cognominal_ has joined #parrot % HG` has joined #parrot % mire__ has left mire__!~Frodo@198-168-222-85.adsl.verat.net % cognominal_ has left cognominal_!~cognomina@82.67.232.89 % slightlyoff has left slightlyoff!~slightlyo@204.14.154.209 % Daveman has joined #parrot % Dave has left Dave!~dave@pool-141-150-16-151.mad.east.verizon.net % wknight8111 has joined #parrot % kj has joined #parrot r25898 | kjs++ | trunk: : [pdd19] : o move the part about = and morphing, so it's before the macro section. : o fix a few pod issues (no errors, but forgotten markup) : o add a note about NCI and short-invocation syntax : o add a note about braced arguments; taken from docs/imcc/macros.pir : o add a note about compilers/pirc/new being able to handle heredoc arguments in macro expansions. diff: http://parrotvm.org/svn/parrot/revision/?rev=25898 % IllvilJa has left IllvilJa!~jilves@emea-netcache1.oracle.co.uk r25899 | kjs++ | trunk: : [pdd19] : add a note about search order of the include directive. diff: http://parrotvm.org/svn/parrot/revision/?rev=25899 % IllvilJa has joined #parrot % wknight8111 has left wknight8111!~nobody@c-71-230-33-251.hsd1.pa.comcast.net r25900 | kjs++ | trunk: : [pirc|heredoc] align assignments and add a comment. No need to regenerate the C file; it's just a matter of layout. diff: http://parrotvm.org/svn/parrot/revision/?rev=25900 r25901 | kjs++ | trunk: : [pirc|macro] add a comment to clarify; it was confusing even to me; and I wrote the thing. : No need to regenerate the C; it's just a comment; nobody should inspect the generated C file. diff: http://parrotvm.org/svn/parrot/revision/?rev=25901 % rdice has joined #parrot % davidfetter has left davidfetter!~davidfett@start.fetter.org % cognominal_ has joined #parrot % skids has left skids!~bri@c-71-233-204-100.hsd1.ma.comcast.net is there a way to say parrot do croak (die with a stack trace) instead of merely dying? % skv_ has joined #parrot % skv has left skv!~skv@87.242.97.68 % skv_ is now known as skv % cognominal_ has left cognominal_!~cognomina@62.210.200.98 % cognominal_ has joined #parrot % cognominal_ has left cognominal_!~cognomina@LAubervilliers-151-13-66-3.w217-128.abo.wanadoo.fr % cognominal_ has joined #parrot % bphillip1 has left bphillip1!~bphillips@nat-003.digitalriver.com r25902 | petdance++ | trunk: : combined similar allocates into a single realloc diff: http://parrotvm.org/svn/parrot/revision/?rev=25902 % AndyAway has left AndyAway!~Andy@64.81.227.163 % rdice has left rdice!~richarddi@CPE001217e365c7-CM00159a01d44c.cpe.net.cable.rogers.com % skids has joined #parrot % gryphon_ has joined #parrot % davidfetter has joined #parrot % marmic has left marmic!~chatzilla@89-253-66-101.customers.ownit.se % stephen has left #parrot good morning, world OH HAI hey pm % marmic has joined #parrot quick, everyone check in their 6KLOC diffs! "on a really productive day, i removed 1000 lines of code" --kernighan cognominal_: probably? I know that we probe for a library that facilitates such things (glib?) but I don't know the particulars. davidfetter: projet I'm on, I plan to remove 50% of the code, damnit. =-) Coke, great :) There's PDB_backtrace, which dumps the Parrot-level backtrace. lathos: you're back! I never went away! I just have a bit less time right now. But I am still working on a strings PDD. lathos++ # strings PDD And, who knows, I may be looking at an ARM port. (We can cross compile, right? *cough) % rdice has joined #parrot i believe we've built on arm kjs++ # many NEWS updates ahem. % skids has left skids!bri@charon.clarku.edu r25903 | petergibbs++ | trunk: : Add the JIT core test on TODO for "repeating alarm" test that chromatic forgot. diff: http://parrotvm.org/svn/parrot/revision/?rev=25903 oh, did I miss something? (likely :-) check the $Id$ in NEWS :P oh, I was looking at the commit messages to parrot-porters and the svn diffs ah, you just haven't gotten to the end yet particle++ # because he deserves it anyway, i'm working on a t/dynoplibs/myops.t failure, should have platforms updates soon okay, great. I got a myops.t TODO pass just a moment ago I'll probably do the release a bit later today ah, somebody todo'd my failing test particle: When I say ARM I mean iPhone really. so, i'll fix it for you, too % Andy is now known as alester % alester is now known as Andy_ OK, who's sitting on Andy? * davidfetter looks up the /ghost command you are, you dingbat % Andy_ has left #parrot % Andy_ has joined #parrot % Andy_ has left #parrot % Andy_ has joined #parrot crud, this is tricky. t/dynoplibs/myops.t fails with msvc, and also under jit let's see if i can fix msvc % davidfetter has left davidfetter!~chatzilla@start.fetter.org % Daveman is now known as Dave % cognominal_ has left cognominal_!~cognomina@LNeuilly-152-21-102-165.w193-253.abo.wanadoo.fr % bphillips has joined #parrot % cognominal_ has joined #parrot particle: ping pong for RT#50968 (capturing .*? in perl6doc), what are you using to test this? i'll nopaste okay, great "particle" at 24.19.3.148 pasted "pod6 test file" (35 lines) at http://nopaste.snit.ch/12361 then just run perl6doc.pbc with --target=parse? correct it works for me with (.*?) gah. did you try -G ? no, i didn't will do, after i finish debugging dynoplibs/alarm.t I'm betting it's another -G bug :-) that would make me happy should I post reply to the ticket, or just wait for you to try it later? wait. should be ~10m okay, no problem. % uniejo has left uniejo!~uniejo@langebro.adapt.dk i'm hoping to get regex/token/rule working in rakudo before release it's ~3 lines of code, but haven't had a chance yet I think the current parsing code isn't quite correct at least, it's not the way I envisioned it feel free to beat me to it. whatever gets it done for release it should be instead of i.e., I didn't want to define special :token and :rule adverbs (unless that's also the way that STD.pm is now doing it) AHA! t/dynoplibs/alarm.t test 7 is a GC BUG! yay! dynoplibs/myops.t that is % cognominal_ has left cognominal_!~cognomina@LNeuilly-152-21-102-165.w193-253.abo.wanadoo.fr afk, lunch & errands % mire__ has joined #parrot % jjore has left jjore!~jjore@c-24-16-241-176.hsd1.mn.comcast.net % diakopte1 is now known as diakopter I would love to figure out how to free the memory from all the yyparsing. % DarkWolf84 has joined #parrot % cognominal_ has joined #parrot % parrot-poke has joined #parrot pmichaud: so, you see something like instead of :token? that's what i'm thinking, just want to make sure we're on the same page er, i mean instead of :rule, of course * particle wonders if > works yet % gryphon_ is now known as gryphon % diakopter is now known as diakopte1 % kismet has joined #parrot Andy_: memory from yyparsing: It's my guess that most of it is allocated in the lexer. that much I know. It's just beyond my understanding of the code at this point tofigure where to free it. ok good. :-) right. I don't know either. I don't know where the mapping from instruction names to instruction numbers is done that's very important I think. but IMCC has instruction names/strings all over the place (in the code generation) % Theory has joined #parrot I don't know if interpolation is defined in .... I don't think it is i.e., is the same as , iiuc and yes, I was thinking that rule would be pmichaud: i have that in now, but it's not working (yet) okay. I really don't want to hold up the release for a working regex/token/rule, though "particle" at 24.19.3.148 pasted "towards correcting regex/token/rule" (38 lines) at http://nopaste.snit.ch/12362 i figured i could just pass :ratchet and :sigspace in the 'options' hash to p6regex, but it's not working you can, but be sure to eliminate the colons ah, hrmm. and we may also need to store ratchet/sigspace in the PAST node that comes back the colons are removed already, automatically just like :qq etc I think we may need to store them in the PAST::Node, then and that may require a modification to PCT/PAST/POST to handle yep, i think it does ok, release away! hrmm, i suppose my patch won't hurt anything, and is closer to what we want sure, that'd be great at the moment the important thing is to get regex working we can get token and rule shortly after that from an interactive run: rule abc { a b c }; say $/ if 'abc' ~~ //; abc regex abc { a b c }; say $/ if 'abc' ~~ //; abc token abc { a b c }; say $/ if 'abc' ~~ //; abc looks correct except for rule yep everything is acting like regex right r25904 % bphillips has left bphillips!~bphillips@nat-003.digitalriver.com r25904 | particle++ | trunk: : [rakudo] bringing regex/token/rule support closer to correct -- everything still acts like regex diff: http://parrotvm.org/svn/parrot/revision/?rev=25904 % rdice has left rdice!~richard_d@CPE0014bfafbbd5-CM0011e6ecf48a.cpe.net.cable.rogers.com % proq has left #parrot r25905 | particle++ | trunk: : [RELEASE] PLATFORMS updates for msvc diff: http://parrotvm.org/svn/parrot/revision/?rev=25905 % rdice has joined #parrot % rdice has left rdice!~richard_d@CPE0014bfafbbd5-CM0011e6ecf48a.cpe.net.cable.rogers.com % chromatic has joined #parrot % cotto__ has left cotto__!~cotto@tide533.microsoft.com % rdice has joined #parrot % cotto__ has joined #parrot % Ron has joined #parrot % cognominal_ has left cognominal_!~cognomina@195.154.53.136 should failures in PLATFORMS be failed test files or individual tests? Files, I believe. i don't recall what i've done in the past % PacoLinux has joined #parrot rant: where the heck is my smoke? Inside the device. Hit it with a hammer to release it. % mire__ has left mire__!~Frodo@198-168-222-85.adsl.verat.net Coke: you've got the power hooked up backwards. You're running it in smoke-retention mode. hurm. smoke just reports # of individual test failures. close enough. r25906 | coke++ | trunk: : [docs] : Update status for intel mac; compress footnote #'s. diff: http://parrotvm.org/svn/parrot/revision/?rev=25906 particle: if $I0 goto regex_start 92 goto word_start 87 unless $I0 goto word_start if $I0 wasn't true, it's clearly going to be false ;-) look closer % mire__ has joined #parrot Oh! These colorful diffs... :-) * particle builds on centos 5.1 x64 I actually had that ladder there for once I figured out how to pass on the right modifiers or call the Right Thing in PGE. Though I never got around to working out how to go do the Right Thing. Those options don't get passed on at all, though. jonathan: yep, but mine is closer to the right thing :ratchet and :sigspace should be passed along to p6regex that's the only bit that needs fixing now If you can figure out how to do that bit, then it's fine. :-) i'll get there eventually I tried a couple of different things, but none of them were the Right One. Then, I don't understand PGE guts. Yet. yep, pge is a beast. PGE needs optimization. we'll talk about that in two minutes * chromatic wants callgrind-compatible output for PIR chromatic: How much do you think it's PGE that needs it, and how much Parrot? jonathan: we'd need some profiling to tell for sure. PGE is recursive descent, not LR, so there are some definite inefficiencies. If we could find spots in PGE that did too much work, I think we could speed it up notably. Algorithmic optimizations are the best ones. phone I'm flying to Kiev tomorrow for UPW, I'll probably do some Rakudo hacking en route. conference season? duck season! Plus some at the workshop, then I'm around in Holland for 5 days. chromatic: Did you look at the sub introspection stuff with regard to call optimization? Not yet. Making that use arity seems useful. Well, all the other bits (is it named, etc) are there too. is ppd25 up to date? I had a thought last night before I fell asleep to extract lots of smaller functions from the larger functions. I've always looked at inter_call.c and been like, "I'll try to understand this another day". :-) jdv79: anything in draft is suspect. ... which that isn't. whoops! :-) I believe it is mostly up to date; allison pushed that out of draft in the last few months. thanks And did a chunk of implementation work on it too. (alon.... right.) seems -G doesn't help my perl6doc bug :( % jhorwitz has joined #parrot % kjs_ has joined #parrot % kj has left kj!~IceChat7@ip565fd420.direct-adsl.nl I can't build the pdd17pmc branch atm. coke: what fails? fails are rare "coke" at 72.228.52.192 pasted "pdd17pmc" (25 lines) at http://nopaste.snit.ch/12364 (that's with a realclean and no modifications outside of languages/) header problem? % kjs_ is now known as kj try make headerizer see if svn st tells you that it modified something couldn't read 'src/revision.c': No such file or directory at tools/build/headerizer.pl line 362. you have a new Makefile? how new? (I updated, realcleaned, config'd, made. boom.) post realclean. i assume so. hrmm. ok, well try looking for /^headerizer in Makefile Andy_: ping yes wrt headerizer, should it really *die* if a source file isn't found? coke has this problem currently. Coke, that's really weird. There's no Parrot_Class on line 139 of src/oo.c for me. http://blodgettm.blogspot.com/2008/02/corollaries-to-blodgett-first-law.html chromatic's url is at http://xrl.us/bghqu r25907 | particle++ | trunk: : [RELEASE] PLATFORMS updates for linux-x86_64 diff: http://parrotvm.org/svn/parrot/revision/?rev=25907 particle: changing .*? to (.*?) works on my platform with the test input you nopasted earlier. what are you getting? pmichaud: that test has many, many pod paragraphs/blocks/etc hurm. if i change any single .*? to (.*?), i get ... hang on (I just changed the one you mentioned in the ticket) ok, if you change the last one (pod_abbreviated_block) you no longer get =FirstName and =MiddleName in the output if you change all three of them, you get 'No POD found.' ah, it's because converting to a capture actually changes the meaning (inside of token) whether it _should_ or not is a question for TimToady :-) *phew* in PGE at least, token containing (.*?) actually acts like (.*?): as in, a capture in a token returns ... ok. which means that we lose the backtracking whereas a bare .*? will properly backtrack hrmm. that's not right .*: is the default under ratchet, and .*? turns off the : as does .*! correct, but does :ratchet affect groupings? not unless they have alternatives it's not that it's treating .*? like .*: -- it's that (.*?) under :ratchet has an implied ':' after the ) by itself grouping doesn't have any branch points ah, I see what you're saing saying okay, so [ a* | ab ] won't backtrack ? but [ a* ] would ? (all assuming :ratchet) thinking http://www.terrybisson.com/meat.html which is hard right now... (and I should've made those *'s into *?'s ) Oh wait, Coke you're on the PDD 17 branch. chromatic: i think coke knows that Now I know that too. Merging from trunk might help. I know there have been a lot of changes and improvements since the previous merge. Alternately, if Allison moved the Parrot_Class struct definition due to the PMC change.... seems like there's two natural ways to run ratchet, and we could default it either way particle: in the meantime, perhaps try (.*?):! I guess ! could override the : now that ! is always meta but ? wouldn't work as a shorthand, since it means something already i'd like $=[.*?]! will that work? lemme see should work, I think. may need :! instead of just ! No POD found. defaulting to : is probably more useful from an optimization point of view yeah, bare ! isn't specced having :ratchet treat as : seems like the correct approach well, leave it that way for now. I agree that converting .*? to (.*?) causes it to lose backtracking is counter-intuitive (under :ratchet). I don't have an obvious answer. i colud turn ratchet off for that but at least I can claim it's not a PGE bug :-) would it help me? :!ratchet I'm curious to know why $=[.*?]:! didn't work you're curious? me too. trying it now chromatic: this is in pdd17pmc. does that matter? Yeah. That's why I couldn't find that line of code in my version. I was looking at trunk. (catching up) "when is this?" particle: $=(.*?):! works note that we can give a compile-time warning on ratcheted (.*?) specifically because it can never produce anything other than "" was trying to validate pdd17pmc branch before you opened it up to the HLL folk for review. =-) The Parrot_Class declaration is at src/pmc/pmc_class.h:133. My guess is that we need a header include for that. or we could just say that any subgroup that ends with ? is bad that way pmichaud++ for giving me a workaround TimToady++ for clearing up the spec (foo .*?): can only match "foo" ever so if we can warn about that, that makes everything okay :) chromatic: regarding profiling subs: if we do the generic "let me attach another sub to run before this one starts and one after it stops", we can roll our own profiling. (and we need that feature anyway) nap & subclass Sub? insufficient. I need to do this to any sub. (that could be a first step for HLLs, of course.) ok, add ATTRs to Subs in pdd17pmc branch then modify invoke to run them properly what happens with a tailcall? I would imagine you'd have to unroll it if the sub you're leaving had an onexit handler. could we somehow do something magic with invoke, itself? rather than having to wrap subs? (just an idle thought) I was thinking of a very simple runloop that could spit out callgrind-compatible output. http://valgrind.org/docs/manual/cl-format.html afk Ok. that helps with the profiling, but not my feature need. =-) but, first things first. (profiling will help tcl immensely, too.) % Andy_ is now known as AndyAway % Ron has left Ron!rblasch@M1716P015.adsl.highway.telekom.at Hey if we're gonna talk profiling, let's talk tonight some time, and I'll see what I can do under shark, if we have a reasonable target but now I must run. % jhorwitz has left jhorwitz!~chatzilla@pool-68-163-20-15.phil.east.verizon.net (profiling parrot level subs, not c level, if that matters) % kj has left kj!~IceChat7@ip565fd420.direct-adsl.nl I was pondering that it could be useful to be able to have a :pmc(...) adverb for .sub's, specifying what PMC to use for sub. Since in Perl 6, there are different types of subs... Regex, Routine, Code, etc. ah, what subclass of sub or whatever % cotto_ has joined #parrot flavor. gotcha. particle: Yes. Insane, or worth writing to the list? % cotto__ has left cotto__!~cotto@tide503.microsoft.com i'm wondering if .const already works for that .const Routine foo = $P0 the syntax is way different, of course but it seems more general there are questions. like does .const register something in the proper namespace slot? ...like .sub does % IllvilJa has left IllvilJa!~jilves@emea-netcache1.oracle.co.uk % wknight8111 has joined #parrot % Limbic_Region has joined #parrot % Andy has joined #parrot % x has joined #parrot % x has left x!~chatzilla@host86-139-25-133.range86-139.btcentralplus.com % x has joined #parrot % PacoLinux has left PacoLinux!~paco@139.Red-80-36-122.staticIP.rima-tde.net * jonathan sleeps, 5:30am start tomorrow nite % x has left x!~chatzilla@host86-139-25-133.range86-139.btcentralplus.com % wknight8111 has left wknight8111!~nobody@c-71-230-33-251.hsd1.pa.comcast.net % rdice has left rdice!~richard_d@CPE0014bfafbbd5-CM0011e6ecf48a.cpe.net.cable.rogers.com % slightlyoff has joined #parrot % wknight8111 has joined #parrot % Andy has left Andy!~Andy@64.81.227.163 % iblechbot has left iblechbot!~iblechbot@ppp-62-216-196-163.dynamic.mnet-online.de % arcady has joined #parrot % HG` has left HG`!~wells@user-544723e5.wfd84a.dsl.pol.co.uk pmichaud: how goes the release? coke: took a short break, I'll restart in an hour or two we're hoping to catch a glimpse of the eclipse tonight but it's overcast here, so might not have much luck (mostly overcast, that is) ok, that's fine. just wondering if I should hold off on some stuff. bah. tell me again how I can easily find out when a module was added to perl core? There's a module, something corelist. Or you can ask the CPAN bot Which appears to have gone away. lathos: awesome, thank you. lathos++ % Andy has joined #parrot ... we have a lot of things left in lib that are in core. =-) % gryphon has left gryphon!~gryphon@dsl-209-221-185-54.zipcon.net we could rip out 5 more modules from lib based on what's in core as of 5.8.0 % rdice has joined #parrot % kid51 has joined #parrot % particl1 has joined #parrot % particle has left particle!~particle@c-24-19-3-148.hsd1.wa.comcast.net % parrot-poke has left parrot-poke!~mollusk@user-112vvlr.biz.mindspring.com Does anyone know whether we're going to have a release this week? kid51: yes, pmichaud is doing it tonight. k. I just took the ticket about the core modules. I'll wait until after the release to delete them from distro. Okay? I'm working on release now Test::Builder will be tougher. why, exactly, is this release 0.6.0 ? There's a minor API change we have to work around. as opposed to 0.5.x? pmichaud, I thought the PMC changes would make it 0.6.0, but they're not in, so I thought it was 0.5.3. No, but the ticket to which Coke posted noted that our minimum Perl 5 version is 5.8. Hence, anything that was core in 5.8 does not need to be in distro. Correct? okay. the ticket says 0.6.0 but I'll release it as 0.5.3 chromatic: What about Test::Builder? (chromatic's understanding corresponds with mine.) kid51, I removed it a few days ago in my local tree and there were some failures with Parrot::Test. k. then how about Test::Simple/More, Class::Struct, Text::Balanced? Assuming I can get all tests to pass without them in distro, are they good to go? I'm not sure how the Test:: ones will fare until we can clean up the Test::Builder issue. The others should be fine. Okay, I'll work at this in pieces over the weekend. chromatic: you mentioned earlier today that there were some recent patches or tickets for pge/pct? or did I misunderstand? ('recent' meaning...? The past few weeks. http://rt.perl.org/rt3/NoAuth/parrot/Overview.html thanks that helps I've been using it myself to clear up newer stuff and patches I missed. * pmichaud bookmarks that very useful link Y'know, I was going to overhaul something the other day and chromatic +1'd me on it, but I'll be darned if I can remmeber what it was. src/string.c src/string.c is the builtin string, right? should I create there a method as well? purl, forget src/string.c