www.parrot.org/ | 291 RTs left | Next release: 2009-07-21
Set by moderator on 14 July 2009.
00:01 kid51 joined 00:04 mokurai joined
dalek rdinal: 295dd7b | (Ted Reed)++ | (2 files):
Implement << for Arrays; test included.
 00:09
kid51 msg mikehh: When you see this during 'make codetest': manifest_tests fail - # Failed test 'No need to regenerate MANIFEST' -- it simply means that someone added/deleted files without updating MANIFEST and/or MANIFEST.SKIP. Someone needs to run tools/dev/mk_manifest_and_skip.pl. 00:11
purl Message for mikehh stored.
00:12 slavorg joined
kid51 Coke: ping 00:15
00:40 dukeleto joined 00:44 dukeleto_ joined 00:56 elmex_ joined 00:59 davidfetter joined 01:08 mokurai joined
dalek TT #841 created by jkeenan++: t/examples/namespace.t is failing 01:49
rrot: r40128 | jkeenan++ | trunk/docs/dev/fhs.pod:
Removing file as per trac.parrot.org/parrot/ticket/694.
 01:51
01:52 dukeleto joined
dalek TT #694 closed by jkeenan++: docs/dev/fhs.pod: Is this file still relevant? 01:52
kid51 Can anyone run a Smolder report on Win32? 02:04
... or perhaps I should say: Would someone please run a Smolder report on a Win32 system?
Our last report was 7 hours ago. I believe the reasons for many of the test failures observed then have been corrected; would appreciate corroboaration. 02:05
dalek TT #40 closed by jkeenan++: t/pmc/exceptions.t unexpectedly passes 02:09
mikehh kid51: yup - so I gathered - I was just reportin' it 02:15
kid51 mikehh: thanks. 02:28
02:35 janus joined 02:46 chromatic joined 03:00 Theory joined 03:20 donaldh joined 03:46 Zak joined 03:47 allison joined 03:52 Khisanth joined 03:53 Theory_ joined 03:57 szabgab joined
cotto bacek_at_work, pingu 04:05
dalek TT #842 created by pmichaud++: [DEPRECATED] NQP PIR q:to:... 04:33
TT #843 created by pmichaud++: PGE will assume {*} at end of every regex 04:40
rrot: r40129 | pmichaud++ | trunk/DEPRECATED.pod:
Add some deprecations for NQP and PGE.
 04:41
04:42 flh joined 04:45 Andy joined 04:49 eternaleye joined
dalek rrot: r40130 | allison++ | trunk/config/gen/makefiles (2 files):
[build] Use direct substitution instead of a makefile variable for

when assigning to a makefile variable, but requires the space in the call. Resolves TT #700. Thanks to Andy Dougherty for the patch.		 05:21
05:23 flh joined 05:41 chromatic joined 05:47 riffraff joined
GeJ Hello everyone 06:01
FYI, since r40128 (removal of docs/dev/fhs.pod) parrot's configure step will fail at init::manifest. 06:03
the file is listed in MANIFEST on line 451. 06:04
chromatic Let me confirm that (but I'm sure you're right). 06:06
cotto GeJ, fixed 06:07
06:07 uniejo joined
GeJ Also did make html for good measure, and it didn't seem to complain about the missing file. 06:07
dalek rrot: r40131 | cotto++ | trunk/MANIFEST:
[MANIFEST] rerun manifest update script after recent file removal
cotto it's easy to forget the steps involved in adding or removing a file 06:08
GeJ Although it doesn't seem to be happy with docs/book/draft/ch03_pir.pod
cotto, chromatic: thanks 06:09
cotto These are good things to be looking at before a release. 06:10
GeJ running `make fulltest` 06:13
cotto especially on your machine 06:15
GeJ hum... it stopped early. :-/ 06:18
I'm not fluent in Makefile, so could someone be kind enough to explain the difference beteween fulltest and fulltest_all. I have an idea and I would like to know how much I'm wrong about it. 06:21
moritz GeJ: fulltest aborts after the first test block which results in a failure 06:22
GeJ moritz: that was my guess. thanks 06:23
bacek_at_work cotto: pongu 06:28
cotto bacek_at_work, why does HashIteratorKey (and some other code from the branch) return PMCNULL or -1 instead of throwing an exception? 06:29
bacek_at_work cotto: to preserve old behavior...
cotto Won't that make such code fail in potentially unexpected ways? 06:30
bacek_at_work It will. But changing behavior without deprecation notice is really bad idea. 06:31
moritz well, you still have a chance to put one in for 1.4
cotto Now's a really good time. 06:32
cotto feels like a broken record
I've only got one song, but everyone will know what it is.
clock? 06:33
purl cotto: LAX: Thu 11:33pm PDT / CHI: Fri 1:33am CDT / NYC: Fri 2:33am EDT / LON: Fri 7:33am BST / BER: Fri 8:33am CEST / IND: Fri 12:03pm IST / TOK: Fri 3:33pm JST / SYD: Fri 4:33pm EST /
NotFound Play it, Sam.
06:35 barney joined 06:39 japhb joined
bacek_at_work So, if we can put some broad deprecation warning in 1.4 stated something like "All hashes/array/iterators will be revisited" 06:40
cotto yes. Actually the merge should probably have waited until aftr 1.4 since it changes an existing behavior. 06:41
the next best thing is a deprecation notice
I do like having saner code, though. 06:42
bacek_at_work cotto: branch didn't change behavior. 06:43
06:44 szabgab joined
bacek_at_work But I thinking about opening another branch for final cleanup which implies changing of behavior 06:45
06:45 Su-Shee joined
Su-Shee hi 06:45
GeJ wow... servus
bacek_at_work O! Can we put deprectation notice for few VTABLEs?
GeJ forget the 'wow...' part.
bacek_at_work Like "nextkey"
purl GeJ, I didn't have anything matching 'wow...' part
cotto bacek_at_work, I'd run it by chromatic or The Architect but it makes sense with your refactoring work. 06:47
chromatic An email to the list would be great.
bacek_at_work I'll try to compose it later tonight. After investigating TGE failure in Lua. 06:48
cotto Mmmm. Legacy. 06:50
cotto goes to sleep 06:55
good night
07:14 mokurai left 07:19 TiMBuS joined 07:20 donaldh joined 07:37 payload joined 08:02 HG` joined 08:05 TiMBuS joined 08:37 payload joined
mikehh 1 failure - examples_tests FAIL - All Others PASS (pre/post config, smolder, fulltest) at r40131 - Ubuntu 9.04 amd64 08:49
t/examples/namespace.t FAILs - this test seems to rely on hash order 08:50
09:03 abesapien joined
mikehh rakudo builds on parrot r40131 - make test/make spectest PASS 09:04
09:12 payload joined 09:13 payload1 joined 09:14 athomason joined
dalek kudo: 3d5f23c | moritz++ | docs/architecture.svg:
[docs] shiny chart of Rakudo's architecture (svg)
 09:29
09:33 bacek joined 09:40 abesapien joined 09:41 payload joined
dalek rrot: r40132 | bacek++ | trunk (2 files):
[pmc] Add StringIterator.shift_pmc and pop_pmc. fperrad++ for reporting
 10:11
bacek seen fperrad 10:12
purl fperrad was last seen on #parrot 30 days, 13 hours, 27 minutes and 21 seconds ago, saying: Whiteknight, when I encoured the same problem, I send a email (with my IP) to osuosl support [Jun 16 20:40:01 2009]
bacek oh...
mikehh: r40133 should make you happy :) 10:19
dalek rrot: r40133 | bacek++ | trunk/t/examples/namespace.t:
[cage] Fix t/examples/namespace.t to not rely on Hash keys order.
 10:21
10:37 clinton joined 10:40 szabgab joined 10:57 MoC joined
Infinoid happy Friday 11:05
moritz TCIF!
Infinoid thank cotto! 11:06
moritz particle: I took a look at prophet (the offline bug tool you mentioned the other day)... do you know where I can find documentatioin for it? 11:14
11:20 donaldh joined
mikehh bacek: should make us all happy - Big Grin - testing now 11:26
11:30 HG` joined 11:40 whiteknight joined 11:43 s1n joined
mikehh Woo Hoo - All Tests PASS (pre/post config, smolder, fulltest) at r40133 - Ubuntu 9.04 amd64 11:49
whiteknight very nice!
bacek it is not... 11:51
whiteknight ???
bacek EPICLY FAILING to grok Lua failure...
mikehh I really now need to set up some other test platforms on this system - I need to install i386 - and then set it up as a Guest vm on amd64 11:52
bacek mikehh: Did you ever check TapTinder? You can probably combine your effort with mj41's to bring more comprehensive testing for Parrot. 11:54
TapTinder?
purl TapTinder is probably software development tool - taptinder.org . For Parrot project running on tt.perl6.cz/ and reporting build failures to #parrot channel as ttbot.
mikehh I was going to run make cover once I get my i386 setup going 11:55
that's going to take a few hours 11:56
what we need to look at is when we get a failure in our languages, or elsewhere - to put in the appropriate tests 11:57
bacek mikehh: few??? I took whole night on my laptop when I tried it. Consider sharing coverage report with everyone! :)
Who is TGE owner/expert/pumpkin? I'm totally stuck... 11:59
whiteknight bacek: I believe Allison may be the owner ofthat 12:12
or, at least the last-known owner 12:13
bacek *sigh*
moritz it always boils down to the same few people how don't have enough spare time :) 12:14
bacek start thinking about "unmerge" keys_revamp branch.
whiteknight don't do that 12:15
What is failing in Lua? 12:16
bacek whiteknight: why? There is some dark parts in Parrot which aren't covered by tests. And some (even Parrot's) "consumers" relying on such parts.
whiteknight: Check mail from fperrad++
* Lua 12:17
Cannot find the attribute 'past' (Lua::Grammar::top on node
Hash) that you asked for.
it's a message from TGE
whiteknight well, that's a problem of needing more tests
is TGE creating any 'Iterator' PMCs directly? 12:19
bacek may be.
But I preserved current behaviour in branch 12:20
ok. In TGE/Tree.pir _scan_node looks very suspicious. 12:24
mikehh rakudo faills 1 test/2 subtests - t/spec/S12-attributes/instance.rakudo - Failed tests: 121-122 12:28
moritz mikehh: when you run that test again, does it fail every time?
mikehh: it has been failing for me once yesterday, but never again since 12:29
nopaste "bacek" at 114.73.164.9 pasted "Hash brokeness in Rakudo" (21 lines) at nopaste.snit.ch/17301 12:38
bacek moritz: it's definitely failing from time to time...
ok. I have to sleep on all this failures. 12:55
bacek must sleep
purl $bacek->sleep(8 * 3600);
bacek See you tomorrow
mikehh moritz: it passed the previous time I ran the trests 12:59
moritz: I just ran it ./perl6 t/spec/S12-attributes/instance.rakudo - it passed the first time then failed 13:03
dalek TT #844 created by gerd++: [PATCH] building Rakudo from an installed Parrot 13:07
mikehh moritz: i ran it about 6 times and it kept failing - I then ran make t/spec/S12-attributes/instance.t - it failed 13:08
moritz mikehh: ok, thanks
mikehh I then ran ./perl6 t/spec/S12-attributes/instance.rakudo again and it passed 4 times before failing
maoritz: I have just run it 10 times - passed 6 failed 4 13:12
moritz: same failures 121-122 13:13
sorry sa/maorizt/moritz/ 13:14
whatever
13:22 TiMBuS joined
Coke wonders if there's a way to ask IRSSI to display the # of lines queued up after the -- more -- prompt. 13:27
13:28 szabgab joined
Coke "will be revisited" is too vague a deprecation notice. 13:29
13:41 skids joined 13:43 dukeleto joined 13:45 HG` joined 13:56 baest joined
dalek rrot: r40134 | fperrad++ | trunk/src/pmc (5 files):
[cage] a lot of consting
 13:56
14:01 baest joined, PerlPilot joined
dalek rrot: r40135 | fperrad++ | trunk/src/pmc/hash.pmc:
[cage] fix warning
 14:23
rrot: r40136 | particle++ | trunk/src/pmc/hash.pmc:
[cage] use best practice for pmc null comparison, fix whitespace for '=' alignment
 14:37
particle omg! i committed something!
Coke given a copy of parrot_test_run.tar.gz, how can I upload this? 14:41
Coke guesses that TT #845 is the answer. 14:48
dalek TT #845 created by coke++: separate tool for uploading to smolder 14:50
14:52 ilia joined 15:05 payload joined 15:10 Su-Shee left 15:20 donaldh joined 15:23 hercynium joined 15:31 jan joined 15:42 flh joined 15:57 chromatic joined
Coke there's still no way to subclass a .Sub and use a vtable override on invoke, right? 16:01
(as a way of wrapping a sub)
mikehh All Tests PASS (pre/post config, smolder, fulltest) at r40136 - Ubuntu 9.04 amd64 16:03
16:03 iblechbot joined
Coke wonders how much overhead it would involve if a .Sub subclass HAD-A anon .Sub that it dispatched the invoke to. 16:03
16:04 payload joined
Coke then the sub could do any prep work it wanted, and .tailcall if it didn't care to do postprocessing. 16:04
16:06 mokurai joined
dalek TT #844 closed by pmichaud++: [PATCH] building Rakudo from an installed Parrot 16:07
pmichaud stupid trac... I didn't mean for that ticket to be closed
particle stupid user. 16:09
whiteknight yeah, Trac has a bunch of errors between the seat and the keyboard :)
particle wishes his former employer would stop paying him 16:10
pmichaud Sorry, it's trac that is stupid in this case. I *wanted* to reply to the ticket in email, but trac won't let me do that.
whiteknight oh, nevermind 16:14
16:14 davidius joined
whiteknight (there are plenty of bugs between my chair and my keyboard, but I guess that's a personal problem) 16:14
Coke tcl has a thing where you can import procs from one namespace into another. You can then force a "forget" of the namespace later, or ask where this command originally was defined. I am pondering updating TclProc to proxy these requests, but am wondering if this sounds more like a job for a TclNamespace. 16:15
(which doesn't exist yet.)
pmichaud well, same here. It's just that in this case trac forced me to go out of my way to contribute knowledge to the ticket, and in so doing I managed to click a wrong button that resulted in the wrong thing happening.
Coke ('forget' is un-import.)
pmichaud: I just poked again on TT #49 - it was originally being pushed by Infinoid, I think, but I don't know if we've pinged OSU about it since months ago. 16:20
mikehh rakudo (3d5f23c) builds on parrot r40136 - make test/make spectest(up to r27602) PASS - Ubuntu 9.04 amd64
Coke to answer my own question, I think I need to create a TclNamespace with TclNamespaceEntries 16:23
mikehh t/spec/S12-attributes/instance.rakudo is still intermittent - passes 3 fails twice 121-122
16:25 davidfetter joined
Coke is any language doing export/import yet? 16:40
(namespace has an export_to, but no way that I see to declare /what/ is exportable. I presume that's supposed to be handled by a subclass that overrides that method.) 16:41
nopaste "coke" at 72.228.52.192 pasted "namespace doesn't respect CTX_HLL?" (28 lines) at nopaste.snit.ch/17308 16:50
pmichaud Coke: Rakudo is doing export/import. The exported symbols are held in special (sub)namespaces 16:53
Coke so if you have something in Foo::Bar, the exports are in Foo::Bar::.exports ? 16:54
pmichaud Foo::Bar::EXPORT::ALL
Coke ok. and those are meant to be inspectable by end users? 16:55
pmichaud (other tags are possible, such as Foo::Bar::EXPORT::default)
in the Perl 6 case, yes.
Coke right. tcl's going to want to keep that hidden.
pmichaud essentially I think that each hll will have to decide how it keeps track of exports (more)
Tene++ and others have been working on making that information available via the compiler object 16:56
but in the case of loadable modules, it does mean that the module being loaded needs to register its export information somewhere.
Coke seems like you'd want it in the namespace object.
pmichaud well, as your nopaste shows, getting custom namespace objects in place is currently a problem.
and in the case of cross-hll import+export, it's the compiler object that is likely to have the translation knowledge, instead of the namespace. 16:57
16:57 Psyche^ joined
pmichaud (also, it's easy to get a hold of a compiler object -- not always so easy to get hold of namespace objects) 16:57
chromatic That nopaste looks fixable.
pmichaud anyway, in your nopaste, the problem is that the subs are loaded (into their namespaces) before the :init sub executes
Coke ok, let me switch that to :immediate. 16:58
ah, good catch.
fixed.
pmichaud the problem I see with :immediate, though, is that I don't know that it survives through .pbc generation.
chromatic That should still store the map in PBC. 16:59
pmichaud i.e., if the .pir is pre-compiled to .pbc, I don't know that it will remember the map
chromatic If it doesn't, that's a bug too.
Coke I think I can work with this for now, and move a bunch of my namespace logic into the actual NS object. 17:00
just trying to get NS import, origin, and forget working properly.
nopaste "pmichaud" at 72.181.176.220 pasted "hll map fails with :immediate" (41 lines) at nopaste.snit.ch/17309 17:02
chromatic Lovely. 17:03
Coke I would imagine that either the interpreter should also be saved out (including the hll mapping state), or that :immediate subs should get saved for freezing, or that there should be a :tag that means "immediate, but after compilation to bytecode" 17:04
but I would expect that to work. (my NS example just generates the wrong type when run from PBC, no nasty assertion failures) 17:05
whiteknight or, another option, we all just ignore the failure and work around it
chromatic HLL mapping needs to save STRING names, because there's no guarantee that they'll have the same type number later.
It also needs to save class declarations so it can recreate them.
pmichaud what if we map to an anonymous class? string names won't help us there. 17:06
(I guess we can say "no mapping to anonymous classes")
chromatic Yeah, you need some moniker to recreate the mapping.
nopaste "pmichaud" at 72.181.176.220 pasted "namespace version of failure with :immediate and hll_map" (47 lines) at nopaste.snit.ch/17310 17:07
pmichaud part of the problem is that :immediate subs don't really "save for freezing" 17:08
we need something that means "freeze, but execute immediately upon load" 17:09
where "immediately upon load" means "even before loading the rest of the bytecode"
afk, lunhc 17:12
Coke seems like keep track of all the immediates and freezing them as inits would do the trick.
chromatic That may work as a first step. 17:14
I'm not sure we always know which of them we need to freeze, though.
TimToady feels like there's some needed abstraction missing
PerlJam This sounds like the same thing as why perl 5 got CHECK and UNITCHECK 17:15
(or similar anyway)
chromatic In this case it's not that we need to *run code* during PBC loading, it's that we need to *define and map types*. 17:16
Coke chromatic: yes, but to do that up front we're running code. is it easier to rerun that code or to save the results of the mappings? 17:23
easier/better/makes fluffier puppies.
the problem, I think, is that this state is in the interpreter - are we freezing the interpreter? if so, when we thaw it, what happens to the interpreter we were just using? 17:24
chromatic I'm not sure what the right answer is, but I think it's helpful to keep in mind the specifics of the problem we need to solve.
Coke someone should open a ticket. 17:25
pmichaud the issue is that we don't have a way to define new types short of running code. 17:34
(at least part of the issue) 17:35
Tene Tickets don't squash bugs. Commits squash bugs.
;)
PerlJam pmichaud: and the current mechanism generates :immediate subs to run that code? 17:37
pmichaud PerlJam: no.
PerlJam: the current mechanism doesn't provide appropriate hooks to map types such as namespaces or subs. 17:38
in rakudo, new types are created by :load :init subs
(because :immediate subs get lost in .pbc files) 17:39
Coke partcl's using :load, as we don't compile down to PBC and are always load_bytecode'd atm. 17:41
17:46 dukeleto joined 17:55 MikHel joined 18:14 szabgab joined
dalek kudo: faf91f9 | pmichaud++ | docs/spectest-progress.csv:
spectest-progress.csv update: 415 files, 11802 passing, 2 failing

   S12-attributes/instance.rakudo 121 - Can initialize one attribute based on another (1)
   S12-attributes/instance.rakudo 122 - Can initialize one attribute based on another (2)		 18:28
18:29 soxet joined 18:30 darbelo joined 18:40 theory joined 18:41 Ryan52 joined 19:53 theory joined 19:56 szabgab joined 19:59 Andy joined 20:14 baest joined 20:20 soxet joined 21:09 allison joined 21:16 bacek joined 21:28 dukeleto joined 21:29 dukeleto joined 22:20 Limbic_Region joined
dalek cnum-dynpmcs: r112 | darbelo++ | trunk/aux/decTest/src/pct/ (2 files):
Push the PCT decTest parser forward, it now emits PIR that looks like it might
 22:32
darbelo make clean 22:33
purl make clean is fuxord, no worky
darbelo Agh. Wrong window.
dalek TT #846 created by jkeenan++: Add 'testing' to Component pane in Trac 22:35
22:36 rg1 joined
nnunley pmichaud: ping 22:44
22:45 kid51 joined, payload joined, MoC joined
nnunley I almost have the nqp bootstrap passing all tests. 22:48
Trying to figure out why string_literal doesn't respect excaped literals, and how the pir c<inline:,> is taken advantage of by the current pir version. 22:49
chromatic: The leaks I saw last night were caused by attempting to bind methods to an undefined class. 22:50
22:50 tetragon joined
chromatic That's odd; I wonder how it did that. 22:54
nnunley chromatic: I know what I have to break in nqp in order to get back into that state. I'll see if I can come up with a minimal testcase in pir based on what it emits. 22:56
chromatic That would be great. 22:57
23:06 mikehh joined 23:38 mikehh joined

Powered by Cro and the Raku® Programming Language.

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

Thank you!