Camelia, the Perl 6 bug

IRC log for #parrot, 2009-10-11

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 japhb Infinoid, OK, where is that committed? I don't see it in http://github.com/Infinoid/dalek-p​lugins/tree/master/modules/local/
00:00 shorten japhb's url is at http://xrl.us/bfrh7j
00:01 Infinoid I didn't commit it anywhere
00:02 japhb Infinoid, Oh, the thing you pasted looked like a commit diff ...
00:02 Infinoid oh.  /me =~ s/commit/push/
00:02 nopaste "kid51" at 70.107.16.27 pasted "pcc_reapply: make corevm on Darwin/PPC: 2 build errors" (2 lines) at http://nopaste.snit.ch/18303
00:03 japhb would you mind pushing?  I think I might have some hack left in me today ...
00:04 TiMBuS joined #parrot
00:05 * japhb applies several dozen dB of sonic energy to the task of increasing the hack
00:07 Infinoid pushed to devel branch.
00:07 japhb Would you prefer I fork or add me to your repo as a committer?
00:08 dalek parrot: r41812 | mikehh++ | branches/pcc_reapply/t/pmc/multidispatch.t:
00:08 dalek parrot: remove TODO from passing test in t/pmc/multidispatch.t
00:08 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41812/
00:08 Infinoid I don't mind giving out commit bits.  I just have to look up how
00:08 japhb github ID same as here ( japhb )
00:10 allison Tene: if I'm getting 26 failures on the branch with the fill_returns_from_c_args patch, does that mean I'm causing the failures, or have you not committed your final fix yet?
00:11 Infinoid japhb: you're added
00:11 japhb Infinoid, thanks!
00:12 Infinoid msg darbelo you've got a commit bit on the dalek-plugins repo.  your plugin and test are in the "devel" branch
00:12 purl Message for darbelo stored.
00:12 Tene allison: I didn't say that all of 'make coretest' passed, I said that I could successfully run 'make test'
00:13 allison Tene: okay, I'll assume I'm not causing those failures (or at least not a substantial number of them) and go ahead and commit
00:13 Tene allison: I also get 26 failures on coretest
00:14 allison Tene: excellent
00:14 Tene allison: http://smolder.plusthree.com/app/pu​blic_projects/report_details/28844 is the smolder report for a full 'make test'
00:14 shorten Tene's url is at http://xrl.us/bfrid2
00:14 allison It is really nice to be able to run 'make test'
00:14 Tene we only have 91 failed in all of 'make test'
00:14 mikehh I think I am only getting 22
00:14 Tene mikeh: yes, I commented earlier that I'm seeing failures that you aren't.  I'm curious about which they are.
00:17 allison Tene: I'm getting 86 test failures on 'make test' (Mac OSX)
00:17 mikehh pcc_reapply - smolder_coretest #28846 at r41811 - 6,544 ok, 22 failed, 100 todo, 162 skipped and 0 unexpectedly succeeded
00:18 Tene allison: I have a total of 10,667 tests.  you?
00:19 allison Tene: urg, just overwrote that window with a diff
00:19 Tene What windows?
00:19 purl windows is, like, one hell of an event
00:19 Tene Ah.
00:19 Tene Well, we'll get there.
00:20 mikehh running make smoke now
00:21 Tene mikehh: looks like the difference comes from t/pmc/threads.t
00:21 Tene You pass more of those than I do.
00:21 mikehh I am running with --optimize (plus otherr stuff)
00:27 mikehh pcc_reapply - make smoke #28848 at r41811 - 10,625 ok, 82 failed, 263 todo, 560 skipped and 0 unexpectedly succeeded
00:27 dalek parrot: r41813 | allison++ | branches/pcc_reapply/src/call/args.c:
00:27 dalek parrot: [pcc] Change 'Parrot_pcc_fill_returns_from_c_args' to use the shared
00:27 dalek parrot: 'fill_results' utility function. Required some changes to the arg/param
00:27 dalek parrot: accessor functions, since args and returns don't pass pointers like params and
00:27 dalek parrot: results.
00:27 purl well, results is @results seems to work but no result => displayed I see the links but not the results...
00:27 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41813/
00:28 * mikehh need to take a break for a bit - bbl
00:28 * Coke skips review.
00:31 mikehh actually the smoke was at r41812 - although it reports 41811 - I thought I had reconfigured
00:32 mikehh bbl
00:37 allison Tene: I'm 86/10526
00:38 allison So, I guess some of your failures are skipped tests on Mac OS X
00:38 Tene could be.
00:39 Tene I'm trying to debug the big pct failure right now.
00:41 Tene looks like a bad plan... should work on other stuff first.  NQP isn't working at all.
00:49 allison Tene: t/compilers/pct/complete_workflow.t?
00:51 allison Tene: it's 2am here, off for some sleep, will dive in again tomorrow
00:52 Tene Goodnight, allison
00:53 allison Tene: curiously, the t/pmc/coroutine.t failure survived complete refactors of the argument and return handling code
00:53 allison Tene: must be unrelated to either
00:53 Tene that failure was already there?
00:54 Tene Hmm... need to figure out what to work on next.
00:54 allison Tene: yes, it's been the same failure for a long time
00:54 allison Tene: which is curious with all the changes we've made
00:54 allison Tene: g'night
00:56 allison Tene: (especially odd since I thought the problem there was that we weren't handling the :optional flag on returns)
00:58 bacek We still don't handle :opt_flags on returns
00:59 Tene we *should*... I just tested it like an hour ago.
00:59 Tene bacek: can you provide me with a test case that doesn't work?
00:59 bacek Tene: coroutine_8.pir
00:59 Tene :P
01:00 bacek When we have single optional value while(1) loop exits too early.
01:03 Tene Huh.
01:04 kid51 pcc_reapply branch: make coretest: Darwin/PPC: I get 24 failures.
01:04 kid51 http://smolder.plusthree.com/app/pu​blic_projects/report_details/28845
01:04 shorten kid51's url is at http://xrl.us/bfrijq
01:04 Tene kid51: we can run 'make test' in pcc_reapply now.
01:06 kid51 Tene:  Does that indicate I should run 'make' rather than 'make corevm'?
01:06 Tene kid51: yes.
01:07 Tene bacek: Huh.  You're right.
01:07 kid51 Do you have any idea when we will no longer need to configure with --buildframes=0 on x86?
01:07 bacek Tene: of course :)
01:07 Tene kid51: when someone gets around to either ripping out or fixing the remnants of oldjit.
01:08 Tene might want to try to con WK into it.
01:16 Tene That is to say, I don't feel confident that I could fix it, so I'd only look at it as a last resort.  WK has talked about JIT a lot and has some history with ASM, and iirc he has fixed the oldjit at least once in the past.
01:33 chromatic Mostly it's just updating some calls to get and set parameters.
01:46 Whiteknight eh?
01:46 Whiteknight who is conning me into what?
01:48 Whiteknight nevermind, /me backscrolls
01:48 Whiteknight and now it's time for bed, later
01:49 Tene Whiteknight: update the frame builder inthe pcc branch.
01:49 Whiteknight right, I'll get on that
01:49 Tene Whiteknight: tat is, that's what you're being conned into.
01:49 Whiteknight gotcha
01:49 Whiteknight goodnight
01:49 Tene 'night
01:55 Tene argh NQP is failing in a very weird way...
01:56 Tene claiming that some class doesn't have a 'returns' method.
02:00 japhb Infinoid, pushed to devel branch of dalek-plugins; all gitorious.t tests now pass.
02:02 Tene pmichaud: ping
02:02 Infinoid nice, I didn't know about m//g before
02:03 japhb Infinoid, :-)
02:04 japhb So ... trade you code for a parrot-plumage feed?  ;-)
02:05 Infinoid Sure, it's worth trying out
02:09 nopaste joined #parrot
02:13 dalek joined #parrot
02:14 Infinoid (from dalek's log) parrot-plumagelog gitorious ATOM parser autoloaded.  http://gitorious.org/parrot-plumage.atom
02:14 japhb OK, I'll try a trivial commit
02:14 bacek joined #parrot
02:17 Infinoid seen masak
02:17 purl masak was last seen on #parrot 1 days, 14 hours, 20 minutes and 14 seconds ago, saying: good post-lunch coma, whiteknight.  [Oct  9 11:49:06 2009]
02:25 JimmyZ joined #parrot
02:27 japhb dalek?
02:27 purl somebody said dalek was #parrot's spammy little rss bot or (see: dalek plugins)
02:28 japhb Infinoid, any way to tell dalek to check feeds immediately?
02:29 Infinoid if it were going to generate output, it already would have
02:29 japhb damn
02:29 Infinoid I've got outside-of-bot debug tools that tell me the plugin should be working, but debugging things in dalek itself is kind of a pain
02:30 Infinoid parrot-plumage: 21cfa3b | japhb++ | :
02:30 Infinoid parrot-plumage: [BUILD] Makefile.in: NQP_PBC is dead, long live PARROT_NQP
02:30 Infinoid parrot-plumage: review: http://gitorious.org/parrot-plumage/parrot-plumag​e/commit/21cfa3b689468dde15544de49337060f2efafa7b
02:30 shorten Infinoid's url is at http://xrl.us/bfriwb
02:30 Infinoid oh!  unlike github, gitorious will let us shorten that hash to 21cfa3b
02:32 japhb There's a commented-out warn() that can be changed to print to the log file, that would show the bot's thinking ... though if the output above was from you running it manually, I'm not sure what's wrong.  :-(
02:33 japhb Infinoid, you mean we can shorten the review link that we send to the channel?
02:33 Infinoid yes
02:34 Infinoid That output was from me preemptively setting $self->{not_first_time} in gitoriousparser.pm and running test.pl on it
02:34 Infinoid test.pl is a crocky sort of dalek-emulator.
02:35 Infinoid (and it's in the repo in modules/local/ if you're curious just how crocky)
02:35 janus joined #parrot
02:35 japhb :-)
02:52 japhb Need dinner now.  I pushed another commit just to see if dalek notices this one.
02:56 dalek parrot-plumage: 305f5b5 | japhb++ | :
02:56 dalek parrot-plumage: [META] Update README to reflect requiring parrot_nqp
02:56 dalek parrot-plumage: review: http://gitorious.org/parrot-plumage/parrot-plumag​e/commit/305f5b515df8a8f06c5f2c36cf012845caa13e54
02:57 shorten dalek's url is at http://xrl.us/bfrizc
03:05 coke joined #parrot
03:06 coke seen pmichaud ?
03:06 purl pmichaud was last seen on #parrot 1 days, 8 hours, 56 minutes and 32 seconds ago, saying: NQP doesn't have closures?  It should.  If it doesn't, it will.  [Oct  9 18:01:09 2009]
03:07 Coke seen dukeleto?
03:07 purl dukeleto was last seen on #parrot 1 days, 3 hours, 43 minutes and 3 seconds ago, saying: allison: ok, will include a test for it then  [Oct  9 23:15:46 2009]
03:07 Coke seen infinoid?
03:07 purl infinoid was last seen on #parrot 31 minutes and 50 seconds ago, saying: (and it's in the repo in modules/local/ if you're curious just how crocky)
03:07 Coke Infinoid: ping.
03:07 Tene oh!  This is going to be horrible to track down!
03:08 Tene in trunk, an nqp compiler method gets a PAST::Val object, while in the pcc branch, it gets a NQP::Grammar  object
03:08 Tene This is horrible!  I'm not looking forward to tracking this down at all!
03:08 Tene Exclamation marks!
03:09 Infinoid Coke: hi
03:10 Coke Infinoid: can you add http://github.com/partcl/partcl to the commit trackr?
03:11 bacek Tene: I can say you why it's failing. Key PMC isn't properly handled in branch. So NQP fetching wrong value.
03:12 Tene bacek: magical coding robot indeed... I think I remember seeing some failing key tests?
03:12 bacek Tene: yeap.
03:13 Infinoid Coke: Sure.  I'm surprised it isn't already there; I had to rework my test suite due to the repo move (and due to the link being on the Languages page)
03:14 Coke Infinoid: partcl's in there elsewhere.
03:14 Coke I'm moving the repo.
03:14 Infinoid basically, if the link on Languages is right, I shouldn't have to do anything manually
03:14 Coke Infinoid: aha.
03:15 Coke I'll change that.
03:15 Infinoid it's already there
03:15 Infinoid however, since I don't see any output from github partcl in the channel log, I'll figure out what went wrong
03:17 Coke Infinoid: nothing.
03:17 purl i guess nothing is open or closed or nothing
03:17 Coke I was trying to be proactive, is all, and I see dukeleto beat me to it.
03:17 Coke sorry I didn't follow through before checking.
03:18 Coke s/checking/asking/ whoops
03:19 Coke anyone here familiar with how rakudo/rakudo is setup?
03:29 Tene wtf@ t/pmc/object-meths_22.pir
03:32 dalek rakudo: 7ec926f | (Kyle Hasselbacher)++ | t/spectest.data:
03:32 dalek rakudo: [spectest.data] fix filename typo
03:32 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/7​ec926fcd2954a0dc9a4581de3158e63fd3c9db7
03:32 shorten dalek's url is at http://xrl.us/bfri35
03:34 Tene wtf keys
03:34 purl keys is Wakeman.
03:34 Tene no, keys is wtf
03:34 purl okay, Tene.
03:36 bacek Tene: yum, Keys :). Problem in $I1 = $P1[$S0]
03:37 bacek [$S0] is Key PMC with reference to STRING register.
03:37 bacek and trying to use it in fill_params lead to dark land on undefined behaviour.
03:38 Tene ;_;
03:39 bacek What you have to do is move clone_key_arg calls from fill_params to build_sig_object.
03:39 bacek And uncomment early return in clone_key_arg
03:39 Tene first step 0: understand clone_key_arg
03:40 bacek And check in which Context this functions called. Because you probably don't need to save/restore registers
03:42 bacek Tene: ignore it. I'll fix it few minutes
03:45 Tene Thank you, bacek.  I was feeling lost.
03:46 bacek Tene: how much expected failures in coretest?
03:46 Tene I think it was, like, 24 or 26 or soething.
03:46 Tene there's some number that vary by platform.
03:46 Tene between 20 and 30
03:47 bacek ok
03:48 Tene Yeah, I got 26 and mikey got 22
03:48 Tene mikeh
03:59 bacek mikehh :)
04:00 bacek Tene: fixed. At least object-meth.t passed
04:00 Tene bacek++
04:00 Tene commited yet?
04:01 Tene yes
04:01 bacek r41815
04:02 Tene nqp still fails. :(
04:02 * Tene rebuilds, just in case.
04:03 bacek Just because there is still 12 failing tests in calling.t
04:03 mberends joined #parrot
04:03 dalek parrot: r41814 | bacek++ | branches/pcc_reapply/src/call/args.c:
04:03 dalek parrot: [cage] Pass proper variable into exception
04:03 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41814/
04:03 dalek parrot: r41815 | bacek++ | branches/pcc_reapply/src/call/args.c:
04:03 dalek parrot: Fix passing Keys.
04:03 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41815/
04:06 Tene nqp still gets the wrong class.
04:07 Tene I'll look through the rest of calling.t
04:07 Tene several issues from tailcalls.
04:09 Tene Still don't know how we're going to catch the "no params" and "no returns" cases.
04:23 dalek parrot: r41816 | bacek++ | branches/pcc_reapply/src/call/args.c:
04:23 dalek parrot: Fix passing constant strings as named returns.
04:23 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41816/
05:02 Tene allison: Wait, no, PARROT_ERRORS_PARAM_COUNT_FLAG can't be on by default in trunk... because that .pir file fails when i explicitly turn that flag on with errorson in trunk
05:02 Tene allison: this is demonstrated in t/op/calling_83.pir
05:04 Tene Oh, um, I think that commit was maybe bad.
05:05 dalek parrot: r41817 | tene++ | branches/pcc_reapply/src/call/args.c:
05:05 dalek parrot: [pcc] Fail when building a sig for a duplicate named parameter
05:05 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41817/
05:05 dalek parrot: r41818 | tene++ | branches/pcc_reapply/src/call/args.c:
05:05 dalek parrot: Revert "[pcc] Fail when building a sig for a duplicate named parameter"
05:05 dalek parrot: This was bad and I didn't test it.
05:05 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41818/
05:14 Tene okay, named returns are definitely broken.
05:14 bacek_at_work Tene, not only them :)
05:18 dalek parrot: r41819 | tene++ | branches/pcc_reapply/src/call/args.c:
05:18 dalek parrot: [pcc] Fail when building a sig for a duplicate named parameter and don't crash.
05:18 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41819/
05:24 bacek_at_work Tene, you have to check flag before throwing exception
05:26 Tene bacek_at_work: there's something weird there... that flag is on by default in the branch, but off by default in trunk... but allison thinks that it's on by default in trunk?  but I don't think it is?  I don't quite know what's going on there.
05:30 bacek_at_work Tene, bacek@illusion:~/src/parrot$ grep -2 'by default' docs/pdds/pdd03_calling_conventions.pod
05:30 bacek_at_work If too many or too few values are provided for the given target registers,
05:30 bacek_at_work Parrot by default will throw an exception for C<get_params>, but not for
05:30 bacek_at_work C<get_results>.  This error behavior can be controlled via the C<errorson> and
05:30 bacek_at_work C<errorsoff> opcodes using C<PARROT_ERRORS_PARAM_COUNT_FLAG> for C<get_params>
05:30 chromatic The parameter check warning?  It's off by default in trunk.
05:30 Tene That's what I thought.
05:30 chromatic I enabled it once and deep hurting.
05:32 Tene well, it's enabled by default in the branch, and needs to be turned off.
05:35 treed I know there's a git command that will do a super-duper clean of the working directory.
05:35 treed That is, deleting everything that's not there because of git.
05:35 treed But I can't remember the command.
05:35 treed git clean -f seems to work partially
05:35 treed I guess rake clobber takes care of the rest.
05:37 bacek_at_work treed, git reset --hard HEAD
05:38 Tene that won't delete non-tracked files.
05:39 treed Is Coding Red on planet parrot yet?
05:41 treed Okay, so.
05:41 treed Method 'iterator' not found for invocant of class 'String'
05:41 treed Did that get removed or something?
05:41 treed Because these tests used to pass.
05:42 Tene a string iterator?  what would that do?
05:42 Tene as well, I'd be surprised if there was an iterator method on the string pmc.
05:42 treed There's certainly nothing in my code calling it anyway.
05:43 treed ack iterator only shows one file where I'm calling iterator as a method
05:43 treed And it's on cardinal;Range, which defines it in the same file.
05:43 treed That error message is at the end of a long chain of current instr.: 'parrot;POST;Compiler;pir_children' pc 9676 (src/POST/Compiler.pir:77) and things like it
05:44 treed (or rather, the start, but the end temporally)
05:45 treed but, anyway, I'd expect a string iterator to iterate over the characters in the string, for one
05:45 treed This came up (or something like it), once before
05:45 treed and got fixed
05:46 Tene even then, you wouldn't get it by calling the 'iterator' method on a string.
05:46 Tene you would use the 'iter' opcode.
05:46 treed Sure.
05:46 treed And nowhere in cardinal is anything explicitly calling that method on a string.
05:46 Tene so if you're calling the iterator method on a string, it's not related to that.
05:47 treed There are only two instances of calling an iterator method on anything in cardinal, and they're both a Range calling its own method
05:48 treed There is some code where cardinal;String calls iter self
05:48 treed There's a lot of iter all over the source.
05:49 treed http://gist.github.com/207475
05:49 treed that's the full error output
05:49 treed I get that from any of three test files.
05:49 treed Something's failing in the compilation step.
05:50 mikehh pcc_reapply - make smoke #28858 at r41819 - 10,629 ok, 78 failed, 263 todo, 560 skipped and 0 unexpectedly succeeded
05:50 treed Also, there have been some speedups in Parrot execution in the last month or so.
05:51 treed The precompiled cardinal test suite now only takes 20 seconds to run.
05:51 treed Used to be 28-30
05:51 treed compilation still takes about as long, though
05:51 treed although that 20 seconds number may be off slightly because of the three files not actually being executed
05:51 treed But I doubt those three make that big of a difference.
05:54 treed 12:03 < mikehh> the culprit -> parrot: r40788 | jonathan++ | trunk/compilers/pct/src/PAST/Compiler.pir
05:54 treed That's where mikehh previously tracked down the start of this test failure.
05:55 treed I note that this file *does* call a method 'iterator' in some cases.
05:56 treed iter = node.'iterator'()
05:56 nopaste "mikehh" at 86.178.198.194 pasted "test summary report for pcc_reapply branch at r41819" (38 lines) at http://nopaste.snit.ch/18306
05:56 treed Can you even use iter as a variable name?
05:57 mikehh note that 48 of the failures are from t/compilers/pct/complete_workflow.t
05:57 Tene Yes, although I'll look at you funny.
05:58 Tene mikehh: yes, as I noted earlier, NQP doesn't actually work.
05:58 treed tene: That's in Parrot.
05:58 treed So don't look at me funny.
05:59 Tene treed: I'll have to find some other more-complimentary way of looking at you, then. ;)
05:59 treed Line 410 of compilers/pct/src/PAST/Compiler.pir is where that 'iterator' method call is
05:59 treed Heh. I didn't realize you swung that way.
06:00 treed I wish the vim syntax highlighting for PIR was more reliable.
06:00 treed Sometimes when I'm scrolling around it seems to lose track of what shit is and color it all the same
06:00 Tene ^L
06:00 Tene but, yeah
06:00 treed Seems to have to do with perldoc
06:00 dalek parrot: r41820 | tene++ | branches/pcc_reapply/src/call/args.c:
06:00 dalek parrot: [pcc] Handle named returns more properly.
06:00 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41820/
06:01 treed Could post_children be expecting something other than a String?
06:01 treed Return the POST representation of evaluating all of C<node>'s
06:01 treed children in sequence.
06:01 treed Seems likely, given that description.
06:02 treed But somehow it's getting a string, but only in very rare cases.
06:03 treed actually, the backtrace says that it's in pir_children, not post_children
06:03 treed but that's not even in that file
06:03 treed Oh, I was looking at the wrong file.
06:03 treed >_>
06:04 treed so, line 77
06:04 Tene Okay, I fixed named returns, but NQP is still schizophrenic.  I don't want to step through every sub in the NQP compiler.
06:07 treed Hm.
06:08 treed Seems like pir_children would have to be handing itself a string.
06:08 treed Which would require that one of the children be a string.
06:08 treed Can I even effect what happens at that level in POST?
06:08 treed (I being cardinal)
07:15 mikehh pcc_reapply - make smoke #28862 at r41820 - 10,630 ok, 77 failed, 263 todo, 560 skipped and 0 unexpectedly succeeded
07:16 mikehh note that 46 of the failures are from an early exit in  t/compilers/pct/complete_workflow.t - Parse errors: Bad plan.  You planned 54 tests but ran 8
07:18 mikehh Failed tests:  7-8 then exit
07:23 nopaste "allison" at 77.98.130.30 pasted "Demo code for Tene, chromatic, bacek demonstrating that parameter count errors is *on* be default in trunk" (9 lines) at http://nopaste.snit.ch/18307
07:24 allison Tene: I thought PARROT_ERRORS_PARAM_COUNT_FLAG was off by default in trun too, but running code examples shows otherwise
07:25 fperrad joined #parrot
07:41 allison Tene: running t/op/calling_83.pir in trunk, after deleting the explicit 'errorson', still gives me an error about "duplicate named argument - 'a' not expected"
07:41 Tene Hmm.
07:41 Tene Okay.
07:42 allison What's strange about the branch is it's *not* giving an error about the duplicate parameter
07:42 Tene allison: it does now.
07:42 allison Even with an explicit errorson
07:42 Tene but it's not guarded with the errors flag
07:42 allison Tene: okay, will try rebuilding again
07:43 Tene allison: named returns work in the branch now, but not named flattened returns.
07:43 allison Tene: will have to check trunk, but I think the duplicate arg shouldn't be guarded by the errors flag
07:44 * Tene nods.
07:44 allison Tene: named flattened returns are where you stuck in the exception "named flattened returns don't work yet"?
07:44 Tene Yes.
07:44 Tene Well...
07:44 Tene if you try it, it fails even earlier.
07:44 allison Tene: so, they just don't have an implementation yet
07:45 allison ah, okay, logic error somewhere
07:45 Tene allison: search for "named returns must have a value"
07:45 Tene that's what gets thrown if you try.
07:46 Tene allison: I'm currently stumped by NQP's weird behavior... it ends up getting an object of the wrong class somewhere, which seems like a very bizarre failure mode.  I haven't felt up to tracking it back through NQP by hand to see where it gets confused.
07:46 allison Tene: I just ran 'svn up' and 'make' and t/op/calling_83.pir still doesn't give me a 'duplicate argument" error
07:46 allison Tene: or any error at all
07:48 Tene allison: that's because I commented out that section at the end of fill_params, which seems to have been the wrong thing to do.  I don't have any idea why I thought it was right.
07:48 dalek parrot: r41821 | mikehh++ | branches/pcc_reapply/src/call/args.c:
07:48 dalek parrot: fix codetest failures - unused assert macros, line length and space before parenthesis
07:48 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41821/
07:49 allison Tene: because uncommenting it makes the PGE build fail
07:49 Tene ah
07:49 allison Tene: (I just tried it)
07:49 allison Tene: but, that means we've go a surface error case
07:49 Tene does PGE do an errorsoff?  chromatic says that error checking is off in trunk, and that everything died horribly when he tried turning it on.
07:49 allison Tene: (that is, a surface fix *looks* like the right one, but isn't)
07:49 allison Tene: yes, that's what I thought too
07:50 allison Tene: I was horribly surprised when I ran code examples on trunk and found that it was errorson by default
07:50 allison Tene: I don't know when it changed
07:51 allison Tene: maybe PGE is setting errors off somewhere?
07:51 allison and the branch isn't picking that up properly?
07:51 Tene no, it doesn't use errorsoff.
07:51 Tene I bet pmichaud would know.
07:53 allison Tene: yes
07:56 allison Tene: maybe it's an error in the code to count how many named args have been used?
07:56 allison Tene: like, a named arg is used somewhere but not counted?
07:56 Tene No, the part that was failing in PGE had one named param, and was called with zero named args.
07:57 allison Tene: oh, then the error message is backwards... hmmm...
07:57 allison Tene: was the one param marked optional?
07:57 allison oh, or was it called with enough positional args to fill the named param?
07:58 Tene one positional arg and one positional param.
07:58 Tene lemme check, though.  I really don't trust my memory.
08:00 Tene oh, yes, I misremembered.
08:01 Tene It's *extra* named params passed to a sub that *doesn't* have a named param.
08:01 nopaste "tene" at 24.10.252.130 pasted "like this" (8 lines) at http://nopaste.snit.ch/18308
08:01 Tene *that* example runs in trunk and fails in the branch.
08:03 Tene even with all error flags turned on
08:04 allison Tene: okay, missing params bad, extra params fine (at the moment)
08:04 Tene Right.
08:05 allison Tene: so we'll need to temporarily mock that behaviour in the branch
08:05 mikehh pcc_reapply - make smoke #28864 at r41821 - 10,630 ok, 77 failed, 263 todo, 560 skipped and 0 unexpectedly succeeded
08:07 allison Tene: technically, I'd call that a bug in trunk, but it's apparently a bug someone was relying on
08:07 Tene Yes.
08:10 mikehh of the 77 listed failures 52 are not run (well 50 the 2 tests exit) - t/compilers/pct/complete_workflow.t - Parse errors: Bad plan.  You planned 54 tests but ran 8
08:10 mikehh and t/library/coroutine.t -  Parse errors: Bad plan.  You planned 6 tests but ran 2
08:11 Tene Yes, pct's workflow test can't run because NQP doesn't work at all.
08:12 Tene not sure about coroutine...
08:12 Tene it has to do with optional returns
08:12 mikehh Yes I noticed - nqp_test failed completely when I ran make test before commiting r41821
08:12 mikehh it doesn't run with make smoke
08:16 mikehh anyway codetest passes at the moment
08:16 mikehh need to take the dog out - bbiab
08:17 allison Tene: okay, but how is that different from t/op/calling_83.pir?
08:17 allison trunk complains "too many named arguments"
08:18 Tene allison: you're going to love this.
08:18 Tene allison: it behaves differently in the case of zero named params.
08:18 nopaste "tene" at 24.10.252.130 pasted "this fails in trunk for allison++" (9 lines) at http://nopaste.snit.ch/18309
08:18 Tene so, definitely a bug.
08:18 allison Tene: okay, that's really a bug :)
08:19 Tene but one that PGE seems to be relying on.
08:19 allison Tene: I'm inclined to say we fix PGE
08:19 Tene nodnod
08:19 Tene or we just put the branc off until pm finishes his PGE replacement. ;)
08:20 allison either it adds a named param, or stops passing that extra named arg
08:20 allison Tene: nooooooooooooo! :)
08:22 allison Tene: do you have the info still lying around on where you found the mismatched call and sub in PGE?
08:22 allison Tene: I'll add a :optional param to the sub
08:23 nopaste "tene" at 24.10.252.130 pasted "There's a reason I have my terminal scrollback set at the maximum it will let me" (7 lines) at http://nopaste.snit.ch/18310
08:28 mikehh alison: I use konsole which allows me to set unlimited scrollback
08:31 mikehh mind you at the moment it is using 493.5MiB Virtual Memory and 60.8 MiB resident memory
08:33 Tene use the ram if you have it.
08:34 mikehh Tene: it is - none of the Virtual Memory is on storage media
08:34 allison mikehh: I'm on my Mac at the moment, the Linux laptop died
08:35 allison mikehh: (the replacement is being shipped as we speak)
08:45 mikehh Actually - I have a graphics card I want to install which will reclaim another 512MB - giving me 8GB to play with
08:48 mikehh the on-board graphics on the Asus M4A78 Pro is ok for what I need but my son just got a new one and his old one is still very good
08:49 allison Tene: blech, it's not just one sub in PGE, it's dozens
08:49 allison Tene: basically, any rule that's called may be called with a named "actions" arg
08:49 Tene yeah
08:49 allison Tene: but, only some of them actually accept a named "actions" param
08:50 allison Tene: I'm adding in a more limited version of your error quieting, that only does it if there were no named parameters handled
08:51 Tene nodnod
08:51 Tene I pretty much know what's going on in PGE, and I'm not too concerned about it.  We can deal with it.
08:51 Tene The NQP issue is proving a lot more frustrating to understand.
08:52 Tene After I wake up tomorrow, I'm going to ave to just walk thgough the entire thing.
08:55 allison joined #parrot
08:55 allison Curious, my Mac just rebooted itself
08:55 * allison thinks this isn't her month for hardware
08:56 treed Mine's been crashing a lot lately.
08:56 treed Actually, my work laptop too.
08:56 treed Something about the screensaver was making it freeze.
08:57 bacek joined #parrot
08:57 allison I'm just hoping my Mac lasts until my new Linux laptop arrives, don't know what I'd do without a computer...
08:58 Tene does perl run on your cellphone?
08:58 Tene ;)
08:58 allison Tene: :)
09:13 jrtayloriv joined #parrot
09:15 dalek TT #1103 created by allison++: [CAGE] Optional named parameters must be explicit
09:26 dalek parrot: r41822 | allison++ | trunk/DEPRECATED.pod:
09:26 dalek parrot: [pcc] Add a deprecation notice for optional named parameters that aren't
09:26 dalek parrot: declared.
09:26 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41822/
09:27 bacek allison: why we need deprecation notice for buggy behaviour?
09:27 allison bacek: because it was part of the interface
09:28 bacek allison: so every single bug (or undefined behaviour) requires deprecation cycle?
09:28 allison it's one of those edge cases where the behavior wasn't defined by the spec, so the implementation is the spec
09:28 allison bacek: no
09:29 bacek allison: where is the crossing line?
09:29 allison I'm sure there are hundreds of little bugs with the old implementation we're cleaning up without realizing it
09:29 allison but if it's big enough that it causes a major external system to fail horribly (in a way that's non-trivial to fix), then it's a deprecation item
09:30 allison and, we certainly won't always catch all of those
09:30 allison but we'll try
09:31 bacek It's specced btw.
09:31 bacek If too many or too few values are provided for the given target registers,
09:31 bacek Parrot by default will throw an exception for C<get_params>
09:32 bacek so, it's better to fix bug in PGE now instead of putting more hacks...
09:33 allison bacek: the problem is, there's no way to be sure we caught all of them
09:33 bacek "all of bugs" or "all of usage"?
09:33 allison bacek: I fixed the ones that allowed PGE to compile, but ran across a huge number more in the PGE tests
09:34 allison "all the cases in PGE that need an explicit optional named parameter added to the sub"
09:34 allison because it's one point invoking a dynamically looked up sub object
09:36 bacek Are all this calls in single place?
09:36 allison I can't guarantee that
09:36 allison and even if we got PGE to pass all the tests, we still wouldn't know if we'd caught all of them
09:37 bacek yes
09:37 bacek but proper behaviour is specced.
09:37 allison (there are times when dynamic dispatch is a pain)
09:38 allison yes, it's true, but the practicality of the deprecation policy is that we don't rip the carpet out from under our users
09:38 allison even if it was our mistake that allowed them to wander down the path in the first place
09:40 bacek I think that all our 6 users can adjust their code to be in-line with spec.
09:41 allison that's what we used to say
09:41 allison and the pain it causes is the reason we have deprecation cycles
09:42 dalek parrot: r41823 | allison++ | branches/pcc_reapply/src/call/args.c:
09:42 dalek parrot: [pcc] Turn on named parameter count checking again, with a smaller restriction
09:42 dalek parrot: to duplicate the behavior of an old bug.
09:42 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41823/
09:49 mokurai left #parrot
10:36 masak joined #parrot
10:55 fperrad_ joined #parrot
10:58 quek joined #parrot
11:13 bacek fperrad: ping
11:14 bacek msg fperrad can you confirm that TT#1093 still failing on win32?
11:14 purl Message for fperrad stored.
11:18 joeri joined #parrot
11:23 mikehh pcc_reapply - make smoke #28871 at r41823 - 10,630 ok, 10,631 ok, 76 failed, 263 todo, 560 skipped and 0 unexpectedly succeeded
11:26 kid51 joined #parrot
11:32 Whiteknight joined #parrot
11:33 ruoso joined #parrot
11:36 dalek rakudo: 0a5ec87 | (Solomon Foster)++ | t/spectest.data:
11:36 dalek rakudo: Turn on S06-operator-overloading/workout.t.
11:36 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/0​a5ec873ee58fc6d3edd5ca642d9b4f3a62a1671
11:36 shorten dalek's url is at http://xrl.us/bfrjxm
11:40 mikehh did make fulltest on pcc_reapply - most of the runcore tests look the same except TestS which has a few more errors - will analyse later
11:54 jrtayloriv joined #parrot
11:57 JimmyZ joined #parrot
12:04 dalek rakudo: e5562c9 | util++ | build/Makefile.in:
12:04 dalek rakudo: RT#69684 Fixed parallel build by adding dependency.
12:04 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/e​5562c90e4bc421b562f68f8fa1ddc1e776eab12
12:04 shorten dalek's url is at http://xrl.us/bfrjzd
12:05 kid51 Util ping
12:16 edgar joined #parrot
12:17 JimmyZ 如果想了解 parrot ,可以去那个频道,另外,chatzilla可以打钩,“Open this Channel at startup”
12:17 JimmyZ sorry, wrong channel
12:31 kid51 msg Whiteknight Have you had a chance to take a look at https://trac.parrot.org/parrot/ticket/1067 ? Is this a GC-related problem?
12:31 purl Message for whiteknight stored.
12:35 kid51 msg Coke Any comment on https://trac.parrot.org/pa​rrot/ticket/1068#comment:3 ? Thanks.
12:35 purl Message for coke stored.
12:36 edgar joined #parrot
12:40 iblechbot joined #parrot
12:45 JimmyZ edgar: ;)
12:49 edgar en ?
12:49 purl i heard en was the pronoun for "of them"?
12:49 edgar what JimmyZ?
12:49 purl JimmyZ are you watching this?
12:49 JimmyZ hello?
12:49 purl Raise your hand in the back if you can't hear me.
12:51 edgar reboot my computer just now
13:19 edgar left #parrot
13:20 fperrad ping bacek_at_work
13:20 purl I can't find bacek_at_work in the DNS.
13:20 quek left #parrot
13:23 fperrad msg bacek I confirm TT#1093 still failing on win32 (today seen with r41820)
13:23 purl Message for bacek stored.
13:38 preflex joined #parrot
14:01 Whiteknight kid51: I'll look at it now
14:14 Psyche^ joined #parrot
14:15 whoppix joined #parrot
14:31 tetragon joined #parrot
14:34 payload joined #parrot
14:51 theory joined #parrot
14:52 davidfetter joined #parrot
14:55 jan joined #parrot
15:01 msmatsko joined #parrot
16:02 Util kid51: pong
16:22 kid51 Util:  Earlier I was looking for some git advice, but I eventually figured it out myself and then got a response on another channel.
16:23 mikehh kid51: ref TT #1061 - is that in the autoprops branch
16:25 dalek parrot: r41824 | allison++ | branches/pcc_reapply/src/ops/core.ops:
16:25 dalek parrot: [pcc] Change the 'result_info' op to get signature information for new calls.
16:25 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41824/
16:25 kid51 mikehh:  No:  https://svn.parrot.org/par​rot/branches/library_files
16:26 kid51 ... though the diff posted to TT #1061 may suffice.
16:27 mikehh kid51: sorry was looking in the wrong branch - in the right one now
16:28 * kid51 tries to remember if he had anything to do with the autoprops branch
16:30 mikehh kid51: was getting verry confuzezed - ha probably need some sleep :-}
16:31 kid51 no, autoprops is Util's branch
16:32 kid51 Speaking of branches ....
16:33 kid51 ... for the Nth time, I wonder if there's anyone on Win32 (or a non-symlinkable filesystem) who could look at the tt509_install_files branch.
16:35 mikehh If I ever get time to properly set up kvm or VirtualBox properly on this box (wireless link) I might even install that as a guest
16:38 Util autoprops is for TT#994
17:02 chromatic joined #parrot
17:17 s1n joined #parrot
17:37 nopaste "mikehh" at 86.178.198.194 pasted "pcc_reapply branch - Summary of make fulltest failures at r41823" (66 lines) at http://nopaste.snit.ch/18311
17:49 mikehh pcc_reapply - make smoke #28886 at r41824 - 10,633 ok, 74 failed, 263 todo, 560 skipped and 0 unexpectedly succeeded
17:49 mikehh two more passing tests
17:50 kid51 quite a formidable list of failures there
17:53 mikehh kid51: which one - the fulltest or smoke? - most of the failures come from t/compilers/pct/complete_workflow.t - Parse errors: Bad plan.  You planned 54 tests but ran 8.
17:53 kid51 I was looking at the fulltest
17:54 mikehh most of the cores have the same failures (one or two differences)
17:54 kid51 Given the number of failures, I doubt we can merge this branch before Oct 20.
17:58 mikehh I think I would like to test rakudo, partcl and cardinal at least, maybe some others
17:59 kid51 Well, if you have a generous supply of electrons, please do so :-)
17:59 moritz wouldn't it make sense to wait with that until PGE really works in that branch? (or does it already? it didn't last I looked)
17:59 mikehh but I would like to get most of the rest passing before that - and nqp_test which fails completely
18:00 jonathan If nqp fails to pass its tests, then I wouldn't hold much hope for Rakudo even compiling, let alone running.
18:01 mikehh jonathan - yes
18:08 Tene yes, it looks better than it is because nqp tests don't run as part of 'make smoke'
18:09 kasih joined #parrot
18:09 nopaste "mikehh" at 86.178.198.194 pasted "pcc_reapply branch - nqp_test failures" (14 lines) at http://nopaste.snit.ch/18312
18:09 mokurai joined #parrot
18:10 kasih salam
18:13 dalek parrot: r41825 | allison++ | branches/pcc_reapply/src/call/args.c:
18:13 dalek parrot: [pcc] Modify the return handling logic, don't throw an exception when we run
18:13 dalek parrot: out of returns before results. Instead, continue looping over positional
18:13 dalek parrot: results, so unfilled optional results and opt_flag results get taken care of.
18:13 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41825/
18:14 * allison down to 28 failing tests in pcc_reapply
18:14 allison and that's 'make test' not 'make coretest'
18:14 moritz allison++
18:15 mikehh excellent - let's check that one :-}
18:17 bacek joined #parrot
18:21 moritz and all nqp tests are clean
18:21 moritz time to check rakudo ;-)
18:24 moritz parrot_config --dumb dies on the branch
18:24 moritz erm, dump
18:24 moritz Null PMC access in get_string()
18:24 moritz current instr.: '_main' pc -1 ((unknown file):-1)
18:25 moritz which prevents rakudo from configuring
18:30 rhr joined #parrot
18:33 mikehh pcc_reapply - make smoke #28889 at r41825 - 10,683 ok, 24 failed, 263 todo, 560 skipped and 0 unexpectedly succeeded
18:33 szabgab joined #parrot
18:34 mikehh and nqp_test passes - looking at fulltest now
19:07 nopaste "mikehh" at 86.178.198.194 pasted "pcc_reapply branch - Summary of make fulltest failures at r41825" (66 lines) at http://nopaste.snit.ch/18313
19:11 allison mikehh: useful, thanks!
19:27 nopaste "bacek" at 114.73.167.154 pasted "hack to allow parrot_config to run for moritz++" (13 lines) at http://nopaste.snit.ch/18314
19:29 chromatic joined #parrot
19:30 bacek sigh... We have to put all old Parrot_run_foo_fromc back.
19:30 Whiteknight why?
19:31 bacek Because of deprecation policy?
19:33 Whiteknight so keep them, but rewrite them to use the new PCC system
19:33 Whiteknight but don't let the evil stay
19:41 jonathan What's the Good Way to do that now?
19:41 jonathan We use those in some of the Rakudo dynpmcs, and while me switching won't allow breaking of deprecation policy, I'd be interested to know what the Better Way is.
19:41 chromatic Through Parrot_pcc_invoke_sub_from_c_args, I believe.
19:42 Whiteknight Am I only seeing like 15 failed tests?
19:42 jonathan Does that use varargs, or let me build up and pass along a CallSignature instead?
19:42 Tene Whiteknight: there are still issues with the jit frame stuff, iirc
19:42 chromatic There's a function to invoke from a CallSignature.
19:42 dalek rakudo: 1305255 | moritz++ | t/spectest.data:
19:42 dalek rakudo: we now pass S12-methods/submethods.t
19:42 jonathan chromatic: OK, that's Good To Know.
19:42 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/1​3052550d9ddd57e8180308dd83898e32b8a70b8
19:42 shorten dalek's url is at http://xrl.us/bfrmb3
19:43 chromatic https://trac.parrot.org/parrot/​wiki/CallingConventionsOverview
19:45 jonathan Thanks.
19:46 joeri left #parrot
19:50 dalek nqp-rx: 08a47f4 | pmichaud++ | src/ (3 files):
19:50 dalek nqp-rx: Create Match objects lazily.
19:50 dalek nqp-rx: review: http://github.com/perl6/nqp-rx/commit/0​8a47f47be2ed1d6bdbcef78f443c0badd948824
19:50 shorten dalek's url is at http://xrl.us/bfrmcu
20:03 allison Whiteknight: I have 28, but yes, we're dropping rapidly
20:04 allison bacek: the Parrot_run_foo_fromc functions are not external functions
20:06 allison bacek: the Parrot_call_* functions are (which is why we provided backward compatibility functions)
20:07 allison jonathan: if you want to build up and pass along a CallSignature, use Parrot_pcc_invoke_from_sig_object instead
20:09 jonathan allison: Is a call signature a kind of "one-shot" thing?
20:09 bacek joined #parrot
20:09 jonathan allison: Or can you keep hold of it and re-use it for multiple calls?
20:10 allison jonathan: you could reuse it, it's not destroyed
20:10 allison jonathan: but, you'd have to make sure it's fully reinitialized each time
20:10 jonathan "fully reinitialized"?
20:10 allison jonathan: as in, no lingering data from the previous call to mess up the next one
20:11 bacek "fully reinitialized" is "destroyed old one and created new one".
20:11 allison a fresh PMC for each call is one way to be absolutely sure there's no old data
20:12 chromatic I have 17 failing coretests.
20:12 allison but, it's not the only way
20:12 allison jonathan: you would also have to be careful that you're not reusing the PMC while some outstanding call expects to still have it
20:14 allison jonathan: the CallSignature currently lasts through the subroutine, containing both arguments and returns, so it's not safe to reuse one to make one call within another, or for recursive calls
20:14 kurahaupo joined #parrot
20:16 kurahaupo gperl
20:16 * kurahaupo wonders why IRC doesn't respond to PINE commands :-( *blush*
20:18 dalek parrot: r41826 | chromatic++ | branches/pcc_reapply (5 files):
20:18 dalek parrot: [PCC] Made MultiSub invocation from C slightly less buggy.  This doesn't
20:18 dalek parrot: completely fix t/src/extend.t #16 (RT #41511), but it crashes in a beter way
20:18 dalek parrot: now.
20:18 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41826/
20:19 chromatic Sure, I threw a patch over the wall, but I did some nasty debugging to figure it out... so there.
20:19 * chromatic eats pancakes
20:21 jonathan allison: OK, got it, thanks. :-)
20:27 nopaste "bacek" at 114.73.145.179 pasted "Patch for rakudo to use public API calls instead of internal functions (for johnatan++)" (100 lines) at http://nopaste.snit.ch/18316
20:28 bacek bah!
20:28 bacek jonathan: nopasted patch was for you, not someone else :)
20:28 dalek parrot: r41827 | bacek++ | branches/pcc_reapply/lib/P​arrot/Pmc2c/PCCMETHOD.pm:
20:28 dalek parrot: [cage] Support bare RETURN() in PCCMETHOD.pm
20:28 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41827/
20:28 jonathan bacek: aha, thanks...
20:28 jonathan :-)
20:29 jonathan I hadn't realized the ones I was calling weren't API. :-)
20:30 bacek me too :)
20:31 allison docs/embed.pod defines the API
20:35 jonathan What, I have to read something?!
20:35 jonathan ;-)
20:40 dalek tracwiki: v13 | mikehh++ | CallingConventionsTasklist
20:40 dalek tracwiki: https://trac.parrot.org/parrot/wiki/CallingCo​nventionsTasklist?version=13&amp;action=diff
20:40 shorten dalek's url is at http://xrl.us/bfrmjg
20:51 allison jonathan: well, it is pretty recent, jhorwitz finished it just before 1.4
21:09 bacek yay. t/src/extend passed. And todoed multisub test too.
21:10 dalek parrot: r41828 | bacek++ | branches/pcc_reapply/src (2 files):
21:10 dalek parrot: Update return_flags in extend.c:append_result
21:10 dalek parrot: review: https://trac.parrot.org/parrot/changeset/41828/
21:11 chromatic Nice tag team, bacek.
21:12 bacek chromatic: you can claim this test (and karma) now :)
21:13 chromatic Doing it now.
21:13 purl well, Doing it now. is it paying the bills?  :)
21:13 bacek rakudo still failing to build...
21:17 bacek oookey.
21:17 bacek There is big chunk of code commented out in Continuation.invoke
21:17 bacek it should be updated to new pcc.
21:25 dalek rakudo: d749d9b | jonathan++ | src/pmc/p (2 files):
21:25 dalek rakudo: Apply patch from bacek++ to get us using documented Parrot API functions in place of some non-API ones.
21:25 dalek rakudo: review: http://github.com/rakudo/rakudo/commit/d​749d9b551cf869358a792e03d167ccb9af63f67
21:25 shorten dalek's url is at http://xrl.us/bfrmoz
22:47 TonyC joined #parrot
23:09 patspam joined #parrot
23:14 davidfetter joined #parrot
23:56 rhr joined #parrot

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

Parrot | source cross referenced