Camelia, the Perl 6 bug

IRC log for #parrot, 2011-05-04

Parrot | source cross referenced

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:00 whiteknight I think we can abstract the underlying threading model from HLLs pretty well
00:00 whiteknight that's precisely why I like the idea of a thread pool and dispatchable messages/tasklets
00:00 whiteknight you make tasks for the things in your program, and don't have to worry how many threads there are, or how they are implemented internally
00:01 whiteknight I absolutely do not want to expose posix threads directly
00:01 whiteknight we can build an interface that looks like posix threads on top of our system, but we need an abstraction
00:01 plobsing I absolutely do not want to care about that discussion.
00:02 whiteknight okay. shelved
00:02 _dolmen_ left #parrot
00:02 plobsing the thing about a concurrent GC (as awesome as that sounds), is that we don't have one right now.
00:03 whiteknight we also don't have threads right now to be safe with
00:03 whiteknight your zeromq bindings withstanding
00:03 plobsing and mod_parrot as well?
00:03 whiteknight do we have mod_parrot
00:03 whiteknight ?
00:03 whiteknight we *did* have it
00:03 whiteknight I can't say we do anymore
00:03 plobsing it exists in a state of disrepair, true.
00:04 plobsing but if we were to attempt to recycle or rewrite, it would be mostly pointless without thread-safety.
00:04 whiteknight as Parrot-Instrument demonstrates, if we don't have it in working order, we don't have it
00:05 whiteknight Okay. So step one is a first-pass, best effort implementation of thread safety?
00:05 whiteknight followed by threads, fixes to thread safety, and the great blue yonder?
00:05 plobsing yes, that sounds about right.
00:06 whiteknight I can find no fault there. We need to decide what our thread-safety mechanism is
00:06 whiteknight a big portion of my threading idea was the need to avoid sharing data at the user level. If we go with that, we only need to worry about securing data internally
00:07 whiteknight and if each thread gets its own read-only proxies and it's own interp/thread, we can avoid sharing data between them without messages
00:07 whiteknight so what we need is to secure the GC, and have thread-safe message mailboxes
00:08 whiteknight mailboxes can come from a library, there are plenty to choose from
00:08 whiteknight which leads us to the GC
00:08 whiteknight and maybe the solution is for us to start writing a concurrent GC core right now so we have it
00:09 plobsing we could punt and just throw in a global-allocator-lock for the moment.
00:09 whiteknight we could, but I really hate that idea
00:09 whiteknight it's just one step short of a global interpreter lock
00:09 cosimo joined #parrot
00:09 plobsing but it isn't nearly as bad because it is relinquished much more frequently
00:12 plobsing and not all of parrot's operations allocate. a fair number of them (the majority?) do, but not all.
00:12 whiteknight do we implement our own locks, which is going to require at least a CAS on every platform we support, or do we use a locking library?
00:13 plobsing we already have a lock abstraction
00:13 whiteknight barely.
00:13 plobsing and I'm generally against going lower than the platform-provided threading abstraction. requiring CAS rules out SPARC for example
00:13 plobsing at least according to liblfds
00:13 whiteknight I know what you're trying to say about how we have stuff and we don't need to start from scratch. I'm of the opinion that what we have is worse than nothing
00:16 whiteknight I'm not trying to be dismissive or pessimistic about it, I sincerely believe we can do better with a clean slate
00:20 plobsing I think there are some useful primitives out of which something that isn't quite functional has been built.
00:20 plobsing for example, the platform (pthread/win32) thread stuff is fairly vanilla.
00:20 whiteknight is it worth the effort to pick through the rubble looking for the occasional thing of value, then fixing it and applying it consistently?
00:21 plobsing there is no point rewritting something that will come out nearly identical no matter how many times you delete and rewrite
00:22 davidfetter left #parrot
00:23 plobsing you mentioned earlier that there were several good thread-safe mailbox libraries available. do you have favorites?
00:23 kid51_at_dinner is now known as kid51
00:23 whiteknight not at the moment. I haven't dug into it deeply enough into any of them to pick favorites
00:28 whiteknight My thinking has all been very preliminary so far. I wasn't expecting to do any serious work on threading until fall at the earliest
00:30 plobsing that would be an important decision, and one that can be made early-on
00:30 whiteknight I take it you are more keen to get working on this sooner rather than later?
00:30 plobsing I'm keen on getting *something* working. It doesn't have to be full, user-visisble threads.
00:31 whiteknight ok
00:33 plobsing I'd like to see Parrot useable for web applications for at least one multi-platform webserver. Apache needs threads, mongrel needs zeromq, fastcgi needs concurrency related things like sockets which we've currently implemented piecemeal.
00:34 whiteknight What I really want to do is avoid complete interp clones for each thread. I really would like to avoid that
00:34 whiteknight but I feel like breaking the interp up in the way it really needs to be is going to be a hassle
00:34 whiteknight nd it's going to require a deprecation poin
00:34 whiteknight point
00:34 plobsing once we have a coherent base, I think a lot of things will fall out as one-off projects, implementable by individuals when the motivation strikes them because the design is already there
00:36 whiteknight I like the concept. I'm having trouble getting around my own extreme demotivation to work on our current rotten threads system
00:38 whiteknight I can definitely start researching libraries
00:38 bubaflub joined #parrot
00:38 whiteknight I do have 6model pretty high up on my priorities list, but I can interleave this with that
00:42 plobsing it is important to me (which is why I've brought it up), so I'm willing to do some work on it. But I'm no architect.
00:43 whiteknight you keep saying that. You either underestimate, or misrepresent yourself, methinks
00:44 whiteknight okay. if you dust off the lock abstraction, we can find a way to fill it with working code
00:45 whiteknight apply a handful of locks to critical portions of code, then move forward from there
00:45 whiteknight sound good?
00:45 plobsing sounds good
00:45 whiteknight then we have an accord
00:45 whiteknight :)
00:45 whiteknight if we can find a portable library which implements some concurrency primitives and message passing routines, that would be a great start
00:46 kid51 plobsing: Good news: tt1931-nci-parameters-deprecation has built successfully on Darwin/PPC at commit dc68cb00c10315c30fd5a7a0456a7673839e7e3d
00:46 kid51 make test now running
00:46 kid51 msg cotto Your LFNW slides look much better than first draft
00:46 aloha OK. I'll deliver the message.
00:46 plobsing kid51: that's great news. if I can get the PPC pass, I can get that thing merged
00:47 cotto_work kid51: thanks
00:47 cotto_work I think they're still to wordy.  I also want to try to make it easier to remember what all the various abbreviations are.
00:48 kid51 plobsing: But in view of today's discussion about performance impacts, have you benchmarked this branch+Rakudo vs master+Rakudo?
00:49 plobsing kid51: I have not. I will do so.
00:57 * whiteknight is heading to bed
00:57 whiteknight goodnight
00:58 whiteknight left #parrot
00:59 cotto_work That seems to have gone well.
01:02 atrodo cotto_work> I actually had the pleasure of sitting with Randall Schwartz at the YAPC::NA dinner last year
01:05 cotto_work I got to hang out with him on Saturday evening too.  He's quite approachable.
01:09 cotto_work That's one of the things I like most about conferences.
01:10 * kid51 could have gone out drinking with Randall Schwartz tonight!
01:10 kid51 But, since we were out with a leading French perlmonger last night, I took a pass.
01:10 kid51 plobsing:  Getting one test failure.
01:12 kid51 Hmm: In 'make test', I got this:
01:12 kid51 t/compilers/pct/complete_workflow.t   Failed tests:  43-44 Non-zero exit status: 2
01:12 kid51 Parse errors: Bad plan.  You planned 54 tests but ran 44
01:12 kid51 But, running the file with 'prove', everything passed.
01:13 kid51 http://smolder.parrot.org/app/​projects/report_details/16122
01:19 plobsing this looks like more of that flaky behaviour we saw earlier when you run things under 'make'
01:20 plobsing I'm not sure what your make program is doing, but I suspect it is modifying the environment somehow
01:20 kid51 I have never seen that failure before, on either box
01:21 kid51 And I have never had to peek into that particular test file before, either.
01:21 plobsing src/gc/gc_gms.c:1231: failed assertion '!PObj_on_free_list_TEST(pmc) || !"Resurrecting of dead objects is not supported"'
01:22 plobsing that's the same failed assertion
01:22 kid51 I'm not sure what earlier flaky behavior you are referring to.  Usually, 'make' gives me less problems than 'prove'
01:22 kid51 Hmm, I didn't look at the smolder directly
01:23 plobsing the next test fails because of file not found and the whole test bails because of it. I suspect that test 43 is supposed to create a file.
01:23 dalek parrot/m0-spec: b079ec9 | cotto++ | docs/pdds/draft/pdd32_m0.pod:
01:23 dalek parrot/m0-spec: add an exception-related TODO to the M0 spec
01:23 dalek parrot/m0-spec: review: https://github.com/parrot/parrot/commit/b079ec9bf0
01:23 kid51 The file consists of several blocks, and yes, files are to be created.
01:24 plobsing kid51: the flaky behaviour I was alluding to was your problems building parrot.
01:24 kid51 Ah, yes.  And we've gotten that error in the past as well.
01:25 Andy joined #parrot
01:27 plobsing kid51: what type of make is it? GNU? custom apple?
01:37 * cotto_work decommutes
01:37 woosley joined #parrot
01:42 kid51 GNU Make 3.80
01:42 kid51 Bad news:  When I configure and build with all g++, 'make' fails
01:42 plobsing with?
01:45 nopaste "kid51" at 192.168.1.3 pasted "tt1931-nci-parameters-deprecation branch on Darwin/PPC: all g++ build: 'make' fails" (770 lines) at http://nopaste.snit.ch/43359
01:45 rurban_ joined #parrot
01:47 plobsing again with that failing assertion! can you run the program that failed outside of make to see if it is the make problem again?
01:47 rurban left #parrot
01:48 rurban_ is now known as rurban
01:48 kid51 ./parrot-nqp --target=pir --output=compilers/opsc/ge​n/Ops/Compiler/Grammar.pir compilers/opsc/src/Ops/Compiler/Grammar.pm  completed
01:49 kid51 IIRC, when I then called 'make' to resume the build, I got one more similar failure, treated in the same way.
01:49 kid51 Yes this next failed:
01:49 kid51 ./parrot-nqp --target=pir --output=compilers/opsc/gen/Ops/Trans/C.pir compilers/opsc/src/Ops/Trans/C.pm
01:50 kid51 And that command also succeeded when run just on its own.
01:50 kid51 This is the same version of 'make' I have used for 5 years.
01:51 kid51 Configuring here with g++ foreach of cc, link and ld (my normal thing is gcc for cc, g++ for the others)
01:51 kid51 And now make is continuing onward
01:52 kid51 ... but I spoke too soon
01:52 kid51 Failed with same error here:
01:52 kid51 ./ops2c --dynamic src/dynoplibs/obscure.ops --quiet
01:52 dmalcolm left #parrot
01:53 bubaflub kid51: could it be related to the new GC failing intermittently / or when certain memory conditions are hit?
01:54 kid51 On this machine, 'master' has been building with what is now default GC
01:55 plobsing bubaflub: that wouldn't explain why it consistently fails when run under make and consistently runs when run individually
01:55 kid51 g++ is finding something it doesn't like that gcc apparently tolerates
01:55 bubaflub plobsing: true.
01:55 kid51 So, 'make' completed -- after 3 restarts :-(
01:55 bubaflub kid51: isn't that usually pointer casts and such?  i'm not sure what would cause that assertion failure.  does it always happen in the same file / same place?
01:56 bubaflub wish i could help test but i don't have access to a PPC machine
01:56 kid51 bubaflub:  Over the past week, depending on which patch plobsing has been asking me to test, we've gotten the same assertion failure in different places -- but apparently consistent re location for any one patch
01:57 bubaflub kid51: ok.
01:57 kid51 Am now trying to smolder -- but going to bed soon
01:58 KaeseEs sort of a philosophical question, what's the rationale behind having (rather) high level stuff like print directly as an opcode for m0?
01:59 kid51 KaeseEs: When cotto gets back from cotto_work, he'll probably be able to answer that question
01:59 KaeseEs okey dokey :)
02:04 plobsing kid51: probably not a terribly good long-term plan, but you could try commenting out the assertion and crossing your fingers
02:10 dukeleto ~~
02:13 bubaflub hola dukeleto
02:18 dukeleto bubaflub: wazzup
02:18 bubaflub dukeleto: writing a final paper, listening to the Black Keys.  you?
02:29 kid51 and on this all=g++ attempt, we once again get one file failing in 'make test':
02:29 kid51 http://smolder.parrot.org/app/​projects/tap_stream/16127/309
02:30 kid51 different file -- t/compilers/opsc/06-opsfile.t  -- but same reason:  failed assertion '!PObj_on_free_list_TEST(pmc) || !"Resurrecting of dead objects is not supported"'
02:30 kid51 with that, sleep
02:30 kid51 left #parrot
02:30 dukeleto bubaflub: oooh, Black Keys is good. Trying to backlog from #ps today
02:31 dukeleto blarg
02:31 ShaneC joined #parrot
02:36 atrodo KaeseEs> the print op in m0 is suppose to be temporary, meant for bootstrapping
02:36 plobsing core.pm > core.pir timings (3-run avg) master: 367.73s tt1931: 358.69s
02:36 plobsing conclusion: no significant impact
02:38 bubaflub plobsing: you're referring to your NCI deprecations branch?
02:38 plobsing yes
02:38 bubaflub ok.
02:38 plobsing kid51 asked for benchmarking
02:39 rohit_nsit08 left #parrot
02:39 * dukeleto => self.nom()
02:43 ShaneC left #parrot
02:47 plobsing do we have any other PPC developpers in the Parrot community?
02:48 hudnix left #parrot
02:49 mtk0 left #parrot
02:51 bubaflub plobsing: http://trac.parrot.org/parrot/wiki/ListOfPlatforms lists only dukeleto and kid51
02:51 bubaflub best bet might be to hit the mailing list
02:56 mtk0 joined #parrot
03:00 jrt4__ joined #parrot
03:03 plobsing ping dukeleto
03:05 Andy left #parrot
03:09 dalek parrot/tt1931-nci-parameters-deprecation: 1149c70 | moritz++ | / (2 files):
03:09 dalek parrot/tt1931-nci-parameters-deprecation: Revert "now that PMCNULL is not NULL, we can update function annotations"
03:09 dalek parrot/tt1931-nci-parameters-deprecation:
03:09 dalek parrot/tt1931-nci-parameters-deprecation: This reverts commit c1f77f7eb90041468c3b13f5b48ae2ca1d1c37a5.
03:09 dalek parrot/tt1931-nci-parameters-deprecation: It broke several rakudo spectests, see TT #2104
03:09 dalek parrot/tt1931-nci-parameters-deprecation: review: https://github.com/parrot/parrot/commit/1149c70960
03:09 dalek parrot/tt1931-nci-parameters-deprecation: dc5f4cf | mikehh++ | src/call/pcc.c:
03:09 dalek parrot/tt1931-nci-parameters-deprecation: add missing ASSERT_ARGS
03:09 dalek parrot/tt1931-nci-parameters-deprecation: review: https://github.com/parrot/parrot/commit/dc5f4cf819
03:09 dalek parrot/tt1931-nci-parameters-deprecation: cccbf37 | mikehh++ | src/debug.c:
03:09 dalek parrot/tt1931-nci-parameters-deprecation: add missing ASSERT_ARGS
03:09 dalek parrot/tt1931-nci-parameters-deprecation: review: https://github.com/parrot/parrot/commit/cccbf37e0a
03:09 dalek parrot/tt1931-nci-parameters-deprecation: 8b0397b | mikehh++ | src/debug.c:
03:09 dalek parrot/tt1931-nci-parameters-deprecation: add missing documentation
03:09 dalek parrot/tt1931-nci-parameters-deprecation: review: https://github.com/parrot/parrot/commit/8b0397ba97
03:09 dalek parrot/tt1931-nci-parameters-deprecation: e3ecaf8 | mikehh++ | src/gc/gc_ms.c:
03:09 dalek parrot/tt1931-nci-parameters-deprecation: add missing documentation
03:09 dalek parrot/tt1931-nci-parameters-deprecation: review: https://github.com/parrot/parrot/commit/e3ecaf848a
03:09 dalek parrot/tt1931-nci-parameters-deprecation: 56d903b | plobsing++ | / (57 files):
03:09 dalek parrot/tt1931-nci-parameters-deprecation: Merge branch 'master' into tt1931-nci-parameters-deprecation
03:09 dalek parrot/tt1931-nci-parameters-deprecation: review: https://github.com/parrot/parrot/commit/56d903b31a
03:33 rohit_nsit08 joined #parrot
03:34 rohit_nsit08 left #parrot
03:49 bacek_at_work seen jnthn__
03:49 aloha jnthn__ was last seen in #parrot 8 hours 39 mins ago saying "pmichaud: #phasers is on at the moment, btw :)".
03:50 bacek_at_work msg jnthn__ on which platform are you? (Related to https://gist.github.com/953615 results)
03:50 aloha OK. I'll deliver the message.
04:16 cotto ~~
04:19 cotto KaeseEs, atrodo is correct.  The ops in the "temporary" category only exist to make life easier for implementors.  Once M0 interps start to mature, we'll take them out.
04:19 dukeleto ~~
04:20 KaeseEs is it intended they get moved up to m1/m2/mX or to HLLs entirely (with some sort of 'a syscall goes here!' instr presumably left in lower levels)
04:21 cotto KaeseEs, it'll be done via the ffi ops and calls to Parrot_sprintf or something similar
04:22 KaeseEs ah-ha.
04:23 cotto That'll be the primary mechanism for interacting with any C code, wither internal to Parrot or external.
04:30 bubaflub left #parrot
04:47 cotto seen allison
04:47 aloha allison was last seen in #parrot 4 days 9 hours ago saying "Andy: sure, Ubuntu brain available for query at any time :)".
04:54 dukeleto cotto: what are you hacking on?
04:54 cotto dukeleto, sending out the pds notice
04:54 cotto followed by looking at the rakudo slowdown, followed by thinking about the structure of the M0 interp
04:55 cotto y tu?
04:55 dalek parrot-instrument: fa403a9 | dukeleto++ | LICENSE:
04:55 dalek parrot-instrument: Add a LICENSE
04:55 dalek parrot-instrument: review: https://github.com/parrot/parro​t-instrument/commit/fa403a9bf7
04:55 dukeleto cotto: trying to get back on my M0 wagon
04:56 dukeleto cotto: and dealing with odd git error due to my harddrive wonking out: https://gist.github.com/954763#comments
04:57 dukeleto benabik: you got any clue how to fix https://gist.github.com/954763 ?
04:57 cotto ouch
04:57 cotto dukeleto, ssd?
04:57 dukeleto cotto: nope
04:57 dukeleto cotto: wonky normal harddrive
04:57 dukeleto cotto: but encrypted filesystems could be to blame
04:58 dukeleto cotto: encrypted filesystems do not fail gracefully when hardware goes to wonkytown
04:58 cotto you can always bundle any unpushed commits, nuke and reclone
04:59 dukeleto cotto: yep. but i am crazy enough to want to understand what is going on :)
04:59 cotto your call
04:59 cotto not sure why you'd want to try to understand what looks like a hardware failure though
05:00 dukeleto cotto: well, i feel that there should be a git util to repair this, I guess
05:00 dukeleto cotto: i want lose too much sleep over it
05:00 cotto git fsck?
05:00 dukeleto s/want/won't/
05:01 dukeleto cotto: good idea
05:01 dukeleto cotto: that seems to be doing something. genius!
05:01 dukeleto cotto: almost every other git command was erroring out
05:01 dukeleto cotto: ooh, it spit out even more mysterious junk
05:02 cotto sounds like a win
05:11 benabik dukeleto: It looks like you're missing objects.
05:11 benabik dukeleto: Did you have to kill a fetch recently?
05:12 benabik dukeleto: I'd suggest moving the pack object git was complaining about and trying fsck again.  .git/objects/pack/pack-43a2872bd32​0ca70652c13c47aa933d2cab5cac7.pack might be the corrupt one.
05:12 * benabik is going AFK again.
05:14 cotto aloha, clock?
05:14 aloha cotto: LAX: Tue, 22:14 PDT / CHI: Wed, 00:14 CDT / NYC: Wed, 01:14 EDT / UTC: Wed, 05:14 UTC / LON: Wed, 06:14 BST / BER: Wed, 07:14 CEST / TOK: Wed, 14:14 JST / SYD: Wed, 15:14 EST
05:17 dukeleto benabik: yep, tried that. no avail.
05:18 cotto what's the right rakudo-related list to send the PDS date to, if any?
05:21 dukeleto cotto: perl6-compiler
05:21 dukeleto cotto: possibly perl6-users as well
05:22 woosley1 joined #parrot
05:23 woosley left #parrot
05:31 woosley1 left #parrot
05:35 jsut joined #parrot
05:40 jsut_ left #parrot
05:40 benabik dukeleto: You could try `git prune`
05:41 birdwindupbird joined #parrot
05:50 jsut left #parrot
05:54 woosley joined #parrot
05:56 cotto dukeleto, is it a good idea to build a backup strategy around bup?
05:57 dukeleto cotto: that is what i plan to do, but haven't yet
05:57 dukeleto benabik: git prune gives the same errors :(
05:58 cotto Apparently SSDs are mostly ticking time bombs that will almost certainly self-destruct within two years or so.
06:01 dukeleto cotto: why do you say that ?
06:01 dukeleto cotto: and by self-destruct you mean lose writability and performance?
06:01 dukeleto cotto: because their failure mode is being write-only, not data loss
06:02 cotto http://news.ycombinator.com/item?id=2505700
06:02 cotto ymmv
06:02 dukeleto it always doess
06:02 dukeleto does, even
06:03 dukeleto cotto: lots of noise in that
06:03 dukeleto cotto: not quite scientific ;)
06:04 cotto dukeleto, sure.  It's a good motivation to start regular backups, even if it's a very noisy and limited data set.
06:04 dukeleto cotto: backups are always a good idea
06:05 cotto dukeleto, exactly
06:05 KaeseEs i am more skeptical than normal of anything where jeff atwood is the source
06:06 cotto KaeseEs, yeah.  codinghorror stopped being on my list of rss feeds a while ago
06:12 Andy joined #parrot
06:22 Andy left #parrot
06:24 benabik dukeleto: Try `git repack -Ad && git prune`.  It should put all "good" objects into a single pack and remove the excess.
06:33 UltraDM joined #parrot
06:35 fperrad joined #parrot
06:47 dalek parrot/tt1589_library_path: 98a30cc | dukeleto++ | / (3 files):
06:47 dalek parrot/tt1589_library_path: [t][TT #1589] Apply patch from soh_cah_toa++ with small tweaks and an extra test
06:47 dalek parrot/tt1589_library_path: review: https://github.com/parrot/parrot/commit/98a30ccf34
07:03 dukeleto benabik: alas, the same errors
07:05 theory left #parrot
07:15 cotto dukeleto, what do you think about adding a count of the number of elements in each chunk to the M0 spec?
07:23 particle1 joined #parrot
07:26 particle left #parrot
07:35 mj41 joined #parrot
07:44 spinclad left #parrot
07:44 spinclad joined #parrot
07:56 dalek parrot/m0-spec: 4b9eb13 | cotto++ | docs/pdds/draft/pdd32_m0.pod:
07:56 dalek parrot/m0-spec: add element count to all M0 chunk segments
07:56 dalek parrot/m0-spec: review: https://github.com/parrot/parrot/commit/4b9eb1356f
07:56 dalek parrot/m0-prototype: 29165dc | cotto++ | src/m0/m0_interp.pl:
07:56 dalek parrot/m0-prototype: stub out bytecode loading in the M0 interp
07:56 dalek parrot/m0-prototype: review: https://github.com/parrot/parrot/commit/29165dca7b
07:56 dalek parrot/m0-prototype: 5aad56f | cotto++ | t/m0/m0bgen.t:
07:56 dalek parrot/m0-prototype: implement element count in chunks, clean up code a bit
07:56 dalek parrot/m0-prototype: review: https://github.com/parrot/parrot/commit/5aad56f2ac
08:05 dalek parrot/m0-spec: b971810 | cotto++ | docs/pdds/draft/pdd32_m0.pod:
08:05 dalek parrot/m0-spec: specify a proper magic number, taken from PackFile_Header_validate
08:05 dalek parrot/m0-spec: review: https://github.com/parrot/parrot/commit/b9718101c5
08:05 dalek parrot/m0-prototype: 72889c0 | cotto++ | t/m0/m0bgen.t:
08:05 dalek parrot/m0-prototype: switch to new magic number for generated m0b tests
08:05 dalek parrot/m0-prototype: review: https://github.com/parrot/parrot/commit/72889c0b13
08:06 cotto definitely time for sleep.  dukeleto, comments appreciated
08:07 cotto 'night
08:13 dalek nqp: 625f17d | moritz++ | t/p6regex/rx_backtrack:
08:13 dalek nqp: [t] fix a simple backtrack test
08:13 dalek nqp: review: https://github.com/perl6/nqp/commit/625f17da0c
08:13 dalek nqp: 4b653ba | moritz++ | src/core/testing.pm:
08:13 dalek nqp: implement todo()
08:13 dalek nqp: review: https://github.com/perl6/nqp/commit/4b653ba01a
08:13 dalek nqp: 8fc008a | moritz++ | t/p6regex/01-regex.t:
08:13 dalek nqp: remove outdated comment from 01-regex.t
08:13 dalek nqp: review: https://github.com/perl6/nqp/commit/8fc008a376
08:16 cosimo left #parrot
08:21 bacek ~~
08:27 dalek nqp: de3220e | moritz++ | t/p6regex/01-regex.t (2 files):
08:27 dalek nqp: [t/p6regex] disable broken PIR script for now
08:27 dalek nqp: review: https://github.com/perl6/nqp/commit/de3220ebeb
08:27 dalek nqp: 48d0448 | moritz++ | t/p6regex/01-regex.t:
08:27 dalek nqp: [t/p6eregex] add first shot of regex test runner
08:27 dalek nqp: review: https://github.com/perl6/nqp/commit/48d0448f31
08:37 dalek nqp: b1d4c13 | moritz++ | t/p6regex/01-regex.t:
08:37 dalek nqp: [t/p6regex] start to expand backslash escapes
08:37 dalek nqp: review: https://github.com/perl6/nqp/commit/b1d4c13915
08:49 dalek nqp: 3d43ac9 | moritz++ | t/p6regex/01-regex.t:
08:49 dalek nqp: [t/p6regex] also expand \xBEEF escapes
08:49 dalek nqp: review: https://github.com/perl6/nqp/commit/3d43ac98e4
08:51 jrt4__ left #parrot
09:14 birdwindupbird left #parrot
09:14 birdwindupbird joined #parrot
09:28 dalek nqp: 9a5e02f | moritz++ | t/p6regex/ (3 files):
09:28 dalek nqp: [t/p6regex] update error message wording
09:28 dalek nqp: review: https://github.com/perl6/nqp/commit/9a5e02fa3d
09:35 dalek nqp: 221501e | moritz++ | t/p6regex/01-regex.t:
09:35 dalek nqp: [t/p6regex] skip looping tests for now; emit late plan
09:35 dalek nqp: review: https://github.com/perl6/nqp/commit/221501e537
09:37 birdwindupbird left #parrot
09:38 birdwindupbird joined #parrot
09:38 birdwindupbird left #parrot
09:47 rurban_ joined #parrot
09:49 rurban left #parrot
09:49 rurban_ is now known as rurban
09:56 dalek nqp: f480bab | moritz++ | build/Makefile.in:
09:56 dalek nqp: [build] run p6regex tests with nqp
09:56 dalek nqp: review: https://github.com/perl6/nqp/commit/f480bab769
10:01 jnthn__ ^
10:02 jnthn__ msg bacek Test results came from amd64 box with 8 GB RAM, running Linux.
10:02 aloha OK. I'll deliver the message.
10:02 bacek jnthn__, aloha!
10:02 jnthn__ o/ bacek :)
10:02 jnthn__ Machine CPU info: Intel(R) Core(TM) i7 CPU 920 @ 2.67GHz, 8 cores
10:02 bacek jnthn__, :)
10:03 bacek looks like my investigations were correct:
10:03 bacek 32-bits: 4G is faster than 2G
10:03 bacek 64-bits: 8G is faster than 4G
10:03 jnthn__ Ah. :)
10:03 bacek 4G: 32-bits is faster than 64-bits
10:04 * jnthn__ is happy that somebody other than pmichaud managed to get the results he was seeing
10:05 dalek nqp: 1226591 | moritz++ | build/Makefile.in:
10:05 dalek nqp: [build] use the current nqp for testing, not the installed one
10:05 dalek nqp: review: https://github.com/perl6/nqp/commit/12265913da
10:07 bacek I'm happy that I finally reproduced this slowdown. It's kinda give me ideas where to look for performance
10:08 woosley left #parrot
10:17 fperrad left #parrot
10:18 fperrad joined #parrot
10:36 mtk0 left #parrot
10:43 mtk0 joined #parrot
11:14 dalek nqp: 6b448c9 | moritz++ | build/Makefile.in:
11:14 dalek nqp: [build] fix previous commit. Where did I have my mind?
11:14 dalek nqp: review: https://github.com/perl6/nqp/commit/6b448c90c1
11:15 woosley joined #parrot
11:57 Patterner left #parrot
11:57 Psyche^ joined #parrot
11:57 Psyche^ is now known as Patterner
12:09 preflex left #parrot
12:09 Kovensky joined #parrot
12:12 preflex joined #parrot
12:15 darbelo joined #parrot
12:20 davidfetter joined #parrot
12:23 cotto_work left #parrot
12:31 Patterner left #parrot
12:31 Psyche^ joined #parrot
12:31 Psyche^ is now known as Patterner
12:38 bubaflub joined #parrot
12:49 jsut joined #parrot
13:17 bluescreen joined #parrot
13:24 whiteknight joined #parrot
13:26 whiteknight good morning, #parrot
13:27 pmichaud (SSDs from backlog)  FWIW, both my SSDs (Intel X25M and X15M) seem to be going strong
13:27 bubaflub morning dukeleto
13:29 dalek nqp: 026eb76 | moritz++ | src/HLL/Compiler.pm:
13:29 dalek nqp: start rewriting HLL::Compiler.parse in nqp
13:29 dalek nqp: review: https://github.com/perl6/nqp/commit/026eb767a2
13:29 dalek nqp: d63b1e1 | moritz++ | src/HLL/Compiler.pm:
13:29 dalek nqp: finish rewriting HLL::Compiler.parse in nqp
13:29 dalek nqp: review: https://github.com/perl6/nqp/commit/d63b1e1cca
13:29 dalek nqp: 5359d2b | moritz++ | src/stage0/ (6 files):
13:29 dalek nqp: update bootstrap
13:29 dalek nqp: review: https://github.com/perl6/nqp/commit/5359d2b87a
13:31 hudnix joined #parrot
13:36 dalek nqp: a954a63 | jonathan++ | src/ModuleLoader.pm:
13:36 dalek nqp: Make global merging slightly less dumb.
13:36 dalek nqp: review: https://github.com/perl6/nqp/commit/a954a63d18
13:36 dalek nqp: 4abbac4 | jonathan++ | src/stage0/ (6 files):
13:36 dalek nqp: Update bootstrap.
13:36 dalek nqp: review: https://github.com/perl6/nqp/commit/4abbac405e
14:07 davidfetter left #parrot
14:08 tadzik 'afternoon #parrot
14:15 whiteknight hello tadzik
14:21 benabik morning
14:37 bubaflub left #parrot
14:42 theory joined #parrot
14:42 UltraDM left #parrot
14:49 hercynium joined #parrot
14:58 bubaflub joined #parrot
15:31 dodathome joined #parrot
15:45 dalek nqp: bbd4b88 | jonathan++ | / (4 files):
15:45 dalek nqp: Factor NQP-specific bits out of HLL::SerializationContextBuilder so that we can use it also in Rakudo.
15:45 dalek nqp: review: https://github.com/perl6/nqp/commit/bbd4b886e3
15:45 dalek nqp: 120ff08 | jonathan++ | src/stage0/ (6 files):
15:45 dalek nqp: Update bootstrap.
15:45 dalek nqp: review: https://github.com/perl6/nqp/commit/120ff083db
16:04 woosley left #parrot
16:12 Andy joined #parrot
16:15 Andy left #parrot
16:16 dukeleto bubaflub: mornin'
16:17 rohit_nsit08 joined #parrot
16:39 rohit_nsit08 left #parrot
16:39 dukeleto self.blarg()
17:00 rohit_nsit08 joined #parrot
17:01 whiteknight dukeleto: ping (privmsg)
17:01 mj41 left #parrot
17:19 lucian joined #parrot
17:31 mj41 joined #parrot
17:31 benabik left #parrot
17:32 cotto_work joined #parrot
17:33 cotto_work ~~
17:35 whiteknight hello cotto_work
17:36 cotto_work hio whiteknight
17:40 cotto_work dukeleto: ping
17:47 rohit_nsit08 left #parrot
17:48 rurban_ joined #parrot
17:49 rurban left #parrot
17:49 rurban_ is now known as rurban
17:50 coke_ obra++
17:55 Andy joined #parrot
17:55 Andy left #parrot
17:56 Andy joined #parrot
17:57 benabik joined #parrot
18:00 davidfetter joined #parrot
18:10 rohit_nsit08 joined #parrot
18:11 lucian left #parrot
18:12 rohit_nsit08 left #parrot
18:14 bubaflub left #parrot
18:15 lucian joined #parrot
18:20 ShaneC1 joined #parrot
18:50 bubaflub joined #parrot
18:51 lucian left #parrot
19:02 lucian joined #parrot
19:08 whiteknight http://www.modernperlbooks.com/mt/2011/05/wh​en-you-lack-cheap-and-easy-polymorphism.html
19:10 whiteknight that post could not be more perfectly timed
19:11 cotto_work how so?
19:11 * cotto_work reads
19:11 whiteknight Because I was planning to write a post about upcoming 6model work where I address several of those same points
19:11 cotto_work niec
19:11 cotto_work nice
19:11 whiteknight er, I have most of a post written, I need to revise and publish it
19:11 cotto_work (or niece, if you prefer)
19:12 whiteknight I'm about 99% certain that he isn't talking to a Parrot-centric audience, so much as holding Parrot up as a [bad] example for Perl5 people to learn from
19:12 whiteknight but, we can learn from ourselves too
19:12 cotto_work we'd do well to
19:13 whiteknight the 6model stuff is just the tip of the iceberg. I'm planning a much more grandiose long-term task list for objects and PMCs
19:14 whiteknight making vtables much less expensive and making subclasses much easier to do for built-ins is a key component
19:14 whiteknight Lorito will move a long way in that direction, but I think we need to put down some foundations first
19:14 cotto_work 6model is a foundation.  Deciding what we'll actually do with it will require a lot of planning.
19:16 whiteknight yes, and there is a lot that we both need and want to do
19:16 whiteknight Fixing some of the glaring flaws in our current system is high priority
19:16 whiteknight enabling better flexibility and better features is important too
19:16 cotto_work whiteknight: one of the things dukeleto++ suggested is that we need a set of glossy brochure-type documents that quickly give people a high-level view of Parrot and its various components and plans.
19:16 whiteknight our object system has been nothing if not stagnant and disappointing until now
19:16 whiteknight cotto_work: at one point, I would have suggsted PDDs to fill that role
19:17 whiteknight I'm still not sure what role PDDs are supposed to fill
19:17 cotto_work Something like that for 6model will be necessary if we want PMC design to be an accessible process.
19:17 cotto_work maybe something like docs/glossy
19:19 whiteknight sounds good to me
19:20 atrodo I've always thought PDDs should be design documents.  Like, if we lost all our code, with the PDDs we could recreate parrot
19:21 atrodo But what cotto_work and dukeleto are suggesting is more marketing material
19:21 whiteknight atrodo: We could recreate parrot as it currently stands if about half those PDDs said nothing besides "Do this as poorly as possible"
19:21 whiteknight no, that's harsh
19:21 whiteknight but there are a handful of systems where that is reality
19:21 kid51 joined #parrot
19:23 atrodo whiteknight> I agree.  But that's reality versus theory
19:24 whiteknight atrodo: that's as good an idea for PDDs as any. What PDDs really need is a purpose, any purpose. Once we have that we can rewrite them to actually meet that purpose
19:24 whiteknight because most PDDs are in severe need of a rewrite
19:25 cotto_work PDD0 says it defines the purpose.
19:25 kid51 at least it indicates that chromatic is still *thinking* about Parrot :-)
19:25 whiteknight I think we really need to move a handful of PDDs back into draft/ and start serious rewrites on them
19:26 tadzik did he leave just after the "Rakudo will target multiple VMs" mess?
19:26 whiteknight PDDs 10, 15, 17, 20, 21, 23, 24 and 25 all come to mind as needing major review/revision in the coming months
19:26 cotto_work tadzik: I don't think he left as much as got really busy with other things.
19:26 atrodo my understanding was that he got real busy with a job
19:27 tadzik I see
19:27 whiteknight it probably didn't hurt to take some time and blow off some steam
19:27 whiteknight I would also like to get together with plobsing one day and go over PDDs 13 and 16 with a fine-toothed comb
19:28 kid51 He should link that to planet.parrotcode.org
19:28 whiteknight If we set up the PDDs as the ideal towards which we are constantly striving, that would be a good step. Because then when PDD did not match reality, we would be able to say conclusively that reality is wrong
19:31 atrodo chromatic++
19:32 whiteknight of course, that would require bumping all PDDs back to draft, and only releasing them when they sufficiently describe what we want, not what we have or had or never wanted
19:32 whiteknight in fact, maybe I'll set up a branch to do just that
19:32 cotto_work whiteknight++
19:32 atrodo whiteknight++
19:33 whiteknight cotto_work: don't ++ me just yet. You're going to be on the hook for a lot of the work too!
19:33 whiteknight :)
19:33 cotto_work then cotto++ too
19:33 atrodo cotto++
19:33 cotto_work I'm sure he'll be helpful.
19:33 atrodo he has no choice at this point
19:33 whiteknight we need an architect astronaut to help hand down the lofty designs
19:34 atrodo not being around means being volunteered
19:34 cotto_work atrodo: good point
19:34 atrodo (which is why there is not atrodo_work)
19:34 whiteknight let's volunteer atrodo_work for some stuff too
19:34 whiteknight get his lazy ass in here cleaning things up
19:35 whiteknight :)
19:35 atrodo_work joined #parrot
19:35 * atrodo_work unvolunteers himself
19:36 atrodo_work left #parrot
19:36 cotto_work I guess that's settled.
19:37 kid51 whiteknight: If we set ourselves an objective of having an East Coast Parrot Developers Gathering, then perhaps plobsing could come down from Ontario to attend
19:37 kid51 afk
19:39 whiteknight is that where he lives? I wasn't aware of that
19:39 benabik left #parrot
19:39 whiteknight Where in Ontario?
19:40 cotto_work whitehorse, yukon iirc
19:42 whiteknight yeah, that's nowhere near Ontario or the east coast
19:42 cotto_work or anything, really
19:44 whiteknight with the first million dollars the PaFo takes in, I say we hold a conference somewhere sunny and fly him in for the week
19:44 whiteknight with the second million, we can meet bacek in person and buy him a beer
19:45 whiteknight and with each subsequent million each each get gold-plated laptops
19:45 atrodo whiteknight++ # Fiscal responsibility
19:47 whiteknight if watching celebrities and wallstreet executives on the news has taught me nothing, it's that when you have a lot of money you should waste it in frivolous and border-line criminal ways
19:47 whiteknight and if you can't trust celebrities and wallstreet executives, who can you trust?
19:47 atrodo hobos?
19:50 benabik joined #parrot
19:50 cotto_work at lfnw, dukeleto chatted with Bradley Kuhn about getting Parrot into the Software Freedom Conservancy.  It sounded like it'd be a really nice way to let a pro do the mundane work if PaFo can get its 501(c)(3) status back.
19:55 whiteknight We never were 501(c)(3)
19:55 whiteknight but yes, it's an option we've been looking at
19:55 cotto_work I thought we were but lost it.
19:56 whiteknight there is a grace period between the time when you incorporate and the time when you become 501(c)(3). If you get the status during the grace period, donations prior to that are exempt retroactively
19:56 whiteknight we missed the grace period, so we are not and never were
20:00 benabik left #parrot
20:02 darbelo left #parrot
20:02 darbelo joined #parrot
20:05 hercynium left #parrot
20:05 hercynium joined #parrot
20:08 perlite_ joined #parrot
20:08 utsl left #parrot
20:08 utsl joined #parrot
20:11 perlite left #parrot
20:12 perlite_ is now known as perlite
20:17 hercynium left #parrot
20:23 pmichaud (east-coast parrot gathering):  fwiw, I'm tentatively expecting to attend fosscon in Philadelphia on July 23.  Maybe we could have a meetup then/there.  Also, perhaps there should be a talk about Parrot there.  (http://fosscon.org/)
20:23 pmichaud I can submit one if nobody else thinks they'll be able to do so.
20:24 whiteknight I've been on the fence about it
20:25 whiteknight but if I know other parroteers are heading that way, maybe I'll think harder about it
20:25 pmichaud where's the doodle poll for PDS?
20:26 whiteknight cotto took it down. Sent out results in an email
20:26 pmichaud email says:  "If you're not
20:26 pmichaud sure what time that is for you, check the following table or look at
20:26 pmichaud the doodle and set your timezone:
20:26 pmichaud "
20:26 kid51 (but I think there were some errors in the UTC listed)
20:26 pmichaud so I was wanting to look at the doodle
20:27 pmichaud (to try to tease out the actual UTC/time)
20:27 pmichaud kid51++ for noticing the error :)
20:27 * kid51 actually compounded the error in response ... so we need a follow-up from cotto
20:29 kid51 pmichaud: If you attend on Sat July 23, that will be strong incentive for me to attend as well.
20:29 pmichaud kid51: I'll let you know asap if I'm a confirmed attendee
20:29 pmichaud I'm also planning to stay a day either side of the conference to be able to meet with others
20:29 pmichaud (and if nobody's available to meet, it's my first trip to philly so I'll sightsee :)
20:30 Andy pmichaud: Do you need MX records for rakudo.org?
20:30 kid51 whiteknight lives north of Phila.  former contributors Infinoid and austin_hastings and jhorwitz are in that area; one GSOCer is in northern NJ.
20:30 Andy Is anybody going to be getting mail at that address?
20:30 pmichaud Andy: ...I don't think anyone will be getting mail there.
20:30 Andy I doubt it, too.
20:31 pmichaud if we decide to set up incoming mail there I'll email you about the mx record
20:31 pmichaud of course, it doesn't hurt to go ahead and add the mx record if you wish :)
20:31 cotto_work whiteknight: I didn't take it down.  I marked the date and closed it to further voting.
20:32 cotto_work kid51: the funny thing is that I spent 10 minute just trying to double-check the dates.
20:32 cotto_work pmichaud: http://www.doodle.com/nf5cip2yrt6a448d#table
20:33 jsut_ joined #parrot
20:33 pmichaud cotto_work++ # thanks
20:34 pmichaud doodle says that's 2100 UTC on the 14th  :-)
20:34 whiteknight left #parrot
20:34 cotto_work ignore anything I say and go with the doodle
20:34 cotto_work including this
20:35 cotto_work I
20:35 cotto_work I'll send out a clarification in  a few minutes
20:35 * pmichaud adds the event to his calendar, in case he can make it.
20:36 lucian_ joined #parrot
20:37 jsut left #parrot
20:41 lucian left #parrot
20:41 dodathome left #parrot
20:42 cotto_work sent
20:42 jevin left #parrot
20:44 jevin joined #parrot
21:05 benabik joined #parrot
21:10 benabik left #parrot
21:12 davidfetter left #parrot
21:21 ShaneC1 left #parrot
21:25 fperrad left #parrot
21:26 davidfetter joined #parrot
21:27 kid51 left #parrot
21:37 theory left #parrot
21:48 cotto_work pmichaud: fosscon will look more interesting with you there.  I'll have to look into it more.
21:49 mj41 left #parrot
21:51 cotto_work I've avoided it so far because the conf's site because they don't have any list of proposed talks so I don't have much idea of who's been applying.
21:55 bluescreen left #parrot
21:58 whiteknight joined #parrot
21:59 benabik joined #parrot
22:03 darbelo_ joined #parrot
22:08 darbelo left #parrot
22:11 arnsholt_ left #parrot
22:12 pmichaud bacek: ping
22:12 bacek pmichaud, pong.
22:12 pmichaud do you want me to patch 3.3 or the current master?
22:13 bacek 3.3
22:13 pmichaud okay, will do
22:13 bacek It should apply clearly
22:13 pmichaud will let you know when I have results
22:13 bacek ok, thanks!
22:15 arnsholt joined #parrot
22:15 bacek afk # $dayjob
22:22 bacek pmichaud, jfyi looks like ms2 behave much better on "fragmented memory". Hopefully is_ptr.patch can improve gms
22:23 * bacek is really afk
22:35 Andy left #parrot
22:36 lucian_ left #parrot
22:36 coke_ Seen Andy?
22:36 aloha Andy was last seen in #parrot 2 hours 6 mins ago saying "I doubt it, too.".
22:37 Andy joined #parrot
22:38 coke_ seen mj41?
22:38 aloha mj41 was last seen in #perl6 5 hours 6 mins ago joining the channel.
22:39 coke_ pmichaud: oooh. perhaps I'll ditch yapc and go to fosscon, then.
22:40 coke_ msg mj41 I cannot run my taptinder client anymore. "bad login or password".
22:40 aloha OK. I'll deliver the message.
22:40 coke_ msg mj41 probably been dead for ages, I haven't checked.
22:40 aloha OK. I'll deliver the message.
22:41 Coke Oh, hey, here I am.
22:41 coke_ left #parrot
22:41 bubaflub left #parrot
22:59 Andy left #parrot
23:03 Coke left #parrot
23:06 Coke joined #parrot
23:09 theory joined #parrot
23:11 Coke left #parrot
23:14 Coke joined #parrot
23:19 Coke left #parrot
23:22 Coke joined #parrot
23:32 theory_ joined #parrot
23:35 theory left #parrot
23:35 theory_ is now known as theory
23:48 darbelo_ left #parrot

| Channels | #parrot index | Today | | Search | Google Search | Plain-Text | summary

Parrot | source cross referenced