Parrot 2.11.0 Released | parrot.org | Log: irclog.perlgeek.de/parrot/today | Keep up with GCI
Set by moderator on 5 January 2011.
mikehh whiteknight: I got a TODO pass with test 127, but test 49 fails with an optimized build (Ubuntu 10.10 i386) it passes without --optimize 00:01
00:01 bacek left
whiteknight ok 00:01
mikehh I fixed test 127, but then looking at your commit, saw you had modified test 49 00:02
but the thing is it works fine with g++, but fails with g++ --optimize 00:03
00:05 fbrito joined 00:08 davidfetter joined
mikehh whiteknight: get the same story with gcc, t/pmc/integer.t passes test 49 with gcc but fails with gcc --optimize 00:12
test 127 ok in all cases 00:13
test 49 - # Have: 8.94427190999916, # Want: 4+8i - in optimized builds, but passes without --optimize 00:15
00:18 kid51 joined 00:21 hercynium left 00:23 kid51 is now known as kid51_at_dinner
whiteknight mikehh: Looking over the old code, I think it's always been broken and the test was passing by pure, dumb luck 00:27
the way the old code was, the result value was never being assigned back to $P0. Not explicitly anyway 00:28
Hah! and here's the reason: That test was marked TODO 00:29
davidfetter dukeleto, ping
whiteknight it never did pass before. I untodo'd it because my changes added in the "passes by dumb luck" behavior in the non-optimized build
dalek rrot: e2af51a | Whiteknight++ | t/pmc/integer.t:
remove the test for Integer.i_multiply on Complex. That test had been TODO'd, and I mistakenly untodo'd it. It has never passed as written (which is probably wrong) and the MULTI case that it was designed to test has been removed
 00:32
whiteknight mikehh: fixed and committed
mikehh whiteknight: yeah I saw that, I think I need to look at what is actually happening in the test, see why it was passing
whiteknight it's deleted now. I'm going to look at test #127 to see why it is passing 00:33
mikehh: nevermind, I see now why I never saw that TODO test pass: I don'thave GMP installed on this box
or, any of my boxen 00:34
dukeleto davidfetter: pong
davidfetter dukeleto, looking for a way to get access to my android device
anybody had any luck with the droid2world?
dukeleto davidfetter: never heard of it 00:38
davidfetter it's one of the newer motorola ones. mine's verizon branded
00:40 NotFound_gf left 00:41 dmalcolm left
whiteknight cotto_work: ping 00:43
cotto_work whiteknight: pong
whiteknight cotto_work: I want to make finalize op no longer experimental
we voted on it at #ps a few weeks ago with weak support and no opposition
cotto_work I remember. 00:44
whiteknight In current Parrot, it's essential for non-crashy behavior
Users of it include NQP, Rakudo, Winxed, etc
cotto_work +1
whiteknight Okay, that's all the encouragement I need. I'm marking it not experimental
cotto_work great 00:45
dukeleto +1 00:46
00:46 gbacon left
cotto_work kid51_at_dinner: ping 00:47
00:50 Yuki`N joined 00:51 contingencyplan left 00:54 rfw left 00:55 rfw joined 00:56 kid51_at_dinner is now known as kid51
kid51 cotto_work: pong 00:56
cotto_work kid51: are you happy with the results of the make help gci task? 01:03
kid51 cotto_work: Frankly, not entirely.
I was hoping that there would be at least one comment on each target -- especially the more obscure ones -- even if that comment was only "Required external dependency." 01:04
After all, I don't know what some of those targets are about.
cotto_work kid51: could you give the student some feedback to that effect? There's no need to be shy about saying that a task wasn't completed to your satisfaction. 01:05
kid51 I'm not sure I've seen that student on channel, nor do I have an email address for him.
In melange or in the Trac ticket?
cotto_work either place is fine 01:06
wherever you think he's more likely to see it
I just want to make sure he doesn't get left hanging.
dalek rrot: 02de970 | Whiteknight++ | / (3 files):
the finalize op is no longer experimental
 01:08
kid51 cotto_work: Different subject. Around noon ET today I was able to build NQP and Rakudo and get them to pass their 'make test' targets. 01:09
Is it your belief that that has persisted?
If so, I can close that ticket.
cotto_work kid51: yes. plobsing++ took care of it Tuesday evening. 01:10
Kristaba: do you have any failing tests in your branch? I see several. 01:12
01:13 bacek joined
bacek cotto, ping 01:13
kid51 cotto_work: Now I can't seem to spot that GCI task. Did someone close it?
cotto_work bacek: pong
bacek: more awesomeness?
bacek cotto, i've got strange GC failure in PIRATE+opmap_aware. Parrot VM: PANIC: custom_mark flag set but no custom mark routine defined! 01:14
kid51 Hmm, yes someone closed it.
bacek It's always during mark of PFOpMap 01:15
cotto_work I did, though I'm not sure why.
kid51 Ah, that someone was named cotto.
cotto_work must have been a misfire
Kristaba cotto: Yes, as I write in the github push comment, I have a lot of failing tests in t/compilers/imcc/opt1.t and opt2.t, but they are due to a bug in the imcc optimizer
whiteknight cotto_work: can you take a gander at TT #1515 eventually (no rush) 01:16
it's pretty open-ended. I don't know whether it needs fixing or not
cotto_work rfw: coordinate with Matt to make sure you're not both refactoring out the same bits of fill_params. 01:18
nopaste "cotto" at 192.168.1.3 pasted "failing tests with kristaba's changes' test_status perl tools/dev/nopaste.pl -t failing" (45 lines) at nopaste.snit.ch/27572 01:19
Kristaba O_o
cotto_work goofy title, but the contents are correct
Kristaba I don't have these errors 01:20
bacek cotto, PIRATE exposes "bloody big memory fuckup"...
Kristaba cotto: Wait, I'll do a make fulltest again
cotto: You're sure that you merge my branch in the zeroargs_subs_checking branch? 01:22
Not in master?
dalek TT #1635 closed by whiteknight++: Runloop unrolling for exception handling 01:23
TT #1635: trac.parrot.org/parrot/ticket/1635
TT #833 closed by whiteknight++: Exception from :init sub in PIR compreg causes problems
cotto_work Kristaba: I checked the git log
dalek TT #833: trac.parrot.org/parrot/ticket/833
cotto_work apparently melange is letting students take more than one task now
matt has two active ones
I lie 01:24
it's fbrito
fbrito hello :D
cotto_work Kristaba: just ignore me until I can look at it at home. My brain is fried.
nopaste "Kristaba" at 192.168.1.3 pasted "Failing tests with "make fulltest" on my changes" (10 lines) at nopaste.snit.ch/27573 01:27
kid51 cotto_work: Commented on melange: www.google-melange.com/gci/task/sho...9385113378 01:30
Kristaba: You said that was on 'make fulltest'. Which runcore? 01:31
Hmm: I'm getting a failure on t/compilers/opsc/06-opsfile.t in 'master' 01:32
nopaste "kid51" at 192.168.1.3 pasted "t/compilers/opsc/06-opsfile.t: New failure." (25 lines) at nopaste.snit.ch/27574 01:33
Kristaba kid51: runcore?
My god! With "make test" I have a huge number of test fails! 01:34
What's the problem? fulltest isn't 'better' than test? 01:35
rfw yikes, fill_params is absolutely ginormouos 01:36
cotto_work rfw: yes. hence the gci task 01:38
rfw: If you can't get in contact with Matt, I recommend you drop the task until you can. There's potential for merge conflicts if you don't coordinate. 01:40
rfw ah okay
is he doing it too?
cotto_work yes. I made two tasks and intended to publish the second once the first was done, but it got published prematurely.
rfw oh okay
i don't have anything else to do so i guess i'll be fine just holding this one 01:41
cotto_work I am looking forward to seeing how that turns out.
Kristaba cotto_work: I'll try to fix this tomorrow, it's really too late in France, but if you have any idea of why fulltest (my task description says to run "full test") doesn't check all tests...
cotto_work Kristaba: that's a bug that we need to fix.
go to sleep. I'll make sure something happens.
Kristaba Okay, good (night/day) #parrot 01:42
cotto_work msg cotto It seems that Kristaba's branch passes under fulltest but fails under test. Figure out what's going on.
aloha OK. I'll deliver the message.
01:43 Kristaba left
kid51 whiteknight: ping 01:43
whiteknight pong 01:44
kid51 Is it possible your last commit led to that failure in t/compilers/opsc/06-opsfile.t? Could there be a problem in the test itself? 01:45
whiteknight kid51: possible, yes
I'm not familiar with that test
kid51 Me neither. 01:46
cotto_work I can look when I get home. What's explodey?
whiteknight test 1 in that file fails 01:48
says it can't parse, but is light on details
kid51 I've got it. It's just a counting problem. 01:49
whiteknight I moved finalize_p from experimental.ops to core.ops. I don't know if I missed a step
plobsing that test hardcodes the number of ops it expects. adding/moving/deleting ops will make it go boom. 01:50
whiteknight awesome 01:51
dalek rrot: 76466ef | jkeenan++ | t/compilers/opsc/06-opsfile.t:
With certain ops no longer experimental, we have to change the expected number of ops parsed in certain files.
whiteknight (that test)--
kid51++
dalek rrot: 0be4b23 | Whiteknight++ | / (2 files):
add a new API function to register/unregister a PMC with GC
 01:52
whiteknight Parrot::Embed is hella old and has extremely limited functionality
kid51 plobsing: Thanks for taking care of the problem that was aggravating NQP yesterday.
whiteknight I'm thinking we're going to want to rewrite it
If we move it out of the main repo, I don't know what the deprecation policy says about it
Yuki`N The problem with forking Configure.pl is I can only figure out how to get one fork running which I have a pipe to, so I can only log stdout or stderr. 01:53
I have a preliminary version as a pull request, but that requires redirecting STDERR to STDOUT. 01:54
dalek rrot/opmap_aware_pmcs: 1cda22b | bacek++ | src/pmc/packfilebytecodesegment.pmc:
Remove unused .destroy
 01:55
rrot/opmap_aware_pmcs: 7fcad83 | bacek++ | src/pmc/packfileopmap.pmc:
Call proper vtable on OpLib
rrot/opmap_aware_pmcs: 1a9ca63 | bacek++ | src/pmc/packfileopmap.pmc:
Made .mark more inheritance-friendly
kid51 The GCI task corresponding to Trac TT #1925 is still unclaimed.
Yuki`N: I have to admit that I'm not really in agreement with the thrust of the RFC in the Trac ticket from which your task was created. 01:56
plobsing msg bacek bytecode->op_{info,func}_table are populated by anything that generates packfiles (imcc, pbc load, etc). op_info_table makes it easy to find out anything about the ops without linearly scanning the mapping, and op_func_table is similar but only for finding functions.
aloha OK. I'll deliver the message.
bacek plobsing, ok, thanks. 01:57
kid51 OTOH, Coke doesn't agree with all of my RFCs, either ;-)
bacek plobsing, so, I have to populate them in PackfileBytecode PMC?
dalek rrot: b1e1172 | cotto++ | src/embed/pmc.c:
apostrophe fix
whiteknight blah. I make that mistake all the time 01:58
plobsing bacek: if you want the packfile to work, then yes.
whiteknight and I know the difference between its and it's. My fingers are just stupid about it
bacek plobsing, hm. They do work now without it. I didn't try dynops yet. 01:59
plobsing bacek: they do work without what? the func_table is required for op dispatch and the info_table gets used by various tools. 02:00
bacek plobsing, may be. But my PMC-generated "hello world" do work without it.
check opmap_aware_pmcs branch 02:01
examples/pir/make_hello_pbc.pir
02:01 mtk left 02:03 davidfetter left 02:08 mtk joined 02:09 whiteknight left
kid51 Do we have any guidelines as to which tests need/ought to be written in parrot-nqp? 02:10
(The file I fixed earlier was the first time I ever ventured into that language.)
plobsing bacek: how does that work? the contant ids don't even agree. 02:11
bacek plobsing, no idea :)
plobsing bacek: so long as you are just dumping to pbc, you don't need either of those tables 02:12
but, if you want these to replace the structs, there are users of these tables
bacek plobsing, yes. This is main purpose of those PMCs.
At least for now.
Just generate workable PBC. 02:13
plobsing bacek: what about bypassing Eval?
bacek plobsing, erm. I don't get this question
plobsing bacek: if the packfile pmcs generated runnable bytecode, compilers could bypass the use of IMCC/Eval.pmc to implement runtime eval. 02:14
Eval is nasty
fbrito I am having some problems with set_*_keyed_str on OrderedHash PMC. when I try to test them, set_*_keyed is called instead
02:14 aantn joined
bacek plobsing, it will be next step. For now I just dump PBC. 02:14
plobsing fbrito: there is no PIR opcode that directly calls *_*_keyed_str 02:15
you have to find some other, indirect route of calling them
fbrito ah, thank you :D. I have probably claimed another impossible task, ehhehe 02:16
cotto_work It's not impossible. It just requires a different solution than the obvious one. 02:17
fbrito hm... 02:21
cotto_work either make x_x_keyed dispatch to x_x_keyed_x or drop x_x_keyed_x 02:25
02:30 shortcircuit joined, shortcircuit left
fbrito removing them don't affect the tests, but... 02:36
cotto_work but? 02:39
cotto_work decommutes, hoping for answers when he gets home 02:40
fbrito I don't know... I just don't fell comfortable removing code from such a big project like Parrot :D
02:40 shortcircuit joined
kapace__ fbrito, would it help if it were a small project? :P 02:41
fbrito I think it is easier to understand the impact of code changes in smaller projects :P 02:42
kapace__ fbrito, just make sure that Rakudo and lua-parrot build, and tests pass 02:44
plobsing fbrito: there are instances of use of that vtable by the internals. it is entirely possible that those places never see objects of that type 02:46
use a search tool (editor, grep -r, ack) to find them and consider whether it makes sense for an object of the type under consideration to be used there 02:47
02:48 aantn left
fbrito ok, I will do some research 02:55
02:57 bacek left
dalek rrot/spaces_in_paths: 257785c | jkeenan++ | t/tools/pmc2cutils/04-dump_pmc.t:
'glob() doesn't handle spaces in path well; use 'readdir' instead.
 02:58
cotto ~~ 03:09
03:18 bacek joined 03:21 contingencyplan joined 03:24 bacek left 03:32 bacek joined
bacek cotto, ping. 03:32
03:37 Yuki`N left
cotto bacek, pong 03:38
bacek cotto, there is some very-very strange GC bug with OpMap. Stuff below it just get collected.
I couldn't figure out why...
cotto does it get marked?
bacek cotto, yes, afaiu 03:42
#3 0xb7eab409 in Parrot_PackfileOpMap_get_integer_keyed_str ( 03:43
interp=0x80f8020, _self=0x9d63748, str_key=0x862a050)
at src/pmc/packfileopmap.c:132
132\t return VTABLE_get_integer_keyed(interp, _self, pmc_key);
(gdb) p _self->flags & b_PObj_on_free_list_FLAG
$3 = 0
(gdb) down
#2 0xb7eab048 in Parrot_PackfileOpMap_get_integer_keyed (interp=0x80f8020,
_self=0x9d63748, key=0x9d6d61c) at src/pmc/packfileopmap.c:83
83\t if (VTABLE_exists_keyed(interp, map_cache, key)) {
(gdb) p map_cache->flags & b_PObj_on_free_list_FLAG
$4 = 524288
cotto what explodes it? 03:44
bacek No idea 03:45
cotto building the branch now
bacek Just regular workflow
cotto complete_workflow.t?
bacek no
just test suite of PIRATE
t/03-pbc/03-sugar.t 03:46
afk # flying back to Sydney 03:47
dalek rrot: a976f75 | Coke++ | lib/Parrot/Pmc2c/VTable.pm:
Don't use Exporter; speed up pmc2c.pl

  (??!?)		 03:48
rrot: cf51aae | Coke++ | lib/Parrot/Pmc2c/UtilFunctions.pm:
Remove useless scalar.
03:54 bacek left 04:05 kid51 left 04:22 bacek joined 04:29 bacek left 04:32 bacek joined
bacek cotto, fixed, apparently :) 04:39
dalek rrot/opmap_aware_pmcs: 21ce7a0 | bacek++ | src/pmc/packfilebytecodesegment.pmc:
set custom_mark flag early.
rrot/opmap_aware_pmcs: a7d80c9 | bacek++ | src/pmc/packfileopmap.pmc:
Fix PFOM.mark. Invoking VTABLE_mark isn't sufficient obviousely.
rrot/opmap_aware_pmcs: e933bc3 | bacek++ | src/pmc/packfile (2 files):
Use more idiomatic accessors.
cotto ok 04:41
quick flight
bacek it will depart in about an hour 04:53
dalek rrot: 97d08bb | dukeleto++ | docs/project/git_workflow.pod:
[docs] How to Maintain a Fork on Github
 05:01
bacek cotto, pingy-ping 05:14
dalek r: a5083f5 | bacek++ | src/POST/Compiler.pm:
Fix selection of :main sub.
r: a5083f5 | bacek++ | src/POST/Compiler.pm:
Fix selection of :main sub.
 05:15
bacek hmmm. Double karma?
sorear, dalek reports commit twice. github problem?
msg sorear, dalek reports commit twice. github problem?
aloha OK. I'll deliver the message.
bacek msg cotto Handling CLI is good "medium" GCI task :). E.g. C<./installable_pir --target=pbc -o foo.pbc foo.pir> 05:16
aloha OK. I'll deliver the message.
cotto ok
bacek msg cotto it will require parrot on branch, some tree optimizations and bloody PIRATE :) 05:17
aloha OK. I'll deliver the message.
cotto msg bacek I'm here. You don't have to msg me.
aloha OK. I'll deliver the message.
bacek yes-yes :)
I'm too lazy to remove "msg " after I already typed it.
cotto bacek, why not merge into master? 05:18
bacek I'm still hacking it a bit.
Can be merged anyway. 05:19
Coke t/src/embed.t (Wstat: 0 Tests: 15 Failed: 0) TODO passed: 3
cotto ok 05:20
merge whenever you hit a natural pause 05:22
bacek in progress
half-an-hour before departure
running make test
cotto, #1038? 05:23
as GCI candidate
sorear bacek: transition issue... parrot/pir wasn't removed from the poll list after pushing was configured 05:24
bacek sorear, ah, ok.
I don't mind double karma :)
sorear I've fixed this for parrot/pir
but somebody with admin access to the parrot organization needs to weed duplicates 05:25
bacek cotto, ^^^^
sorear too bad github has no post-receive API
cotto dukeleto has an admin bit for parrot on github 05:26
dalek rrot: e588ed8 | bacek++ | / (12 files):
Merge branch 'opmap_aware_pmcs'
r: 398bca1 | bacek++ | src/POST/Compiler.pm:
Fix handling .return. Claim back one more test.
 05:29
r: 74d2c77 | bacek++ | src/POST/Compiler.pm:
Fix storing Keys in ConstantTable. Claim few more tests back.
 05:32
rrot: 91e3883 | petdance++ | src/ (3 files):
removed unused variables
 05:35
rrot: 2369288 | petdance++ | / (12 files):
Merge branch 'master' of github.com:parrot/parrot
05:38 bacek left
fbrito hm, is anyone here? 05:40
cotto hi
fbrito ah, sorry. figured out my self :) 05:42
05:43 rurban_ joined
cotto ok, but don't let it happen again 05:43
fbrito I am making some tests with OrderedHash and negative index values 05:44
dalek rrot/leto/embed_grant: 2bd6a62 | dukeleto++ | t/src/embed.t:
[t] Add extend_vtable_output_is() to prevent repeated boilerplate in embed tests
 05:45
rrot/leto/embed_grant: c20ee48 | dukeleto++ | t/src/embed.t:
[t] Convert many tests to use extend_vtable_output_is and delete a lot of code
05:45 rurban left, rurban_ is now known as rurban
fbrito for an n-element OrderedHash, hash[-n] = 1 and hash[-(n+1)] = 1 are really supposed to be the same? didn't know about that :o 05:49
cotto yuck 05:51
fbrito if I create an empty hash and push 7 elements to it, hash[-7] = 0 and hash[-8] = 0 are both going to change the 1st value of the hash to 0?! :o 05:54
cotto just cut that out. It's not a sane behavior. 05:55
do I even want to know what hahs[-9] = 0 does?
kapace__ lol
05:57 ppant joined
fbrito we need to test how the code handles with negative index when they are bigger than the number of elements 06:05
s/with// 06:08
06:29 chromatic left 06:32 aantn joined
fbrito I can't understand when is the line 349 be called (tapir2.ro.vutbr.cz/cover/cover-resu...-pmc.html) 06:35
going to be* 06:37
dalek m-eta-wink-kzd: 8df082c | plobsing++ | src/o (2 files):
proper way to check that an object has a method in winxed (emulated in JS too)
 06:46
m-eta-wink-kzd: ac4faa0 | plobsing++ | src/ometa-base.winxed:
implement {g,s}et_int_keyed on OMeta to allow integer state in grammars
m-eta-wink-kzd: e92018c | plobsing++ | src/ (2 files):
use quote-new and init_pmc for JumpTable to get initialization working
m-eta-wink-kzd: 2cfaabc | plobsing++ | src/winxed-compiler.dual:
add quoted-new syntax
m-eta-wink-kzd: 9483502 | plobsing++ | src/Ī©Ī·.capo.winxed:
fix thinko in JumpTable.toTree()
m-eta-wink-kzd: cde831c | plobsing++ | src/Ī©Ī·.coda.winxed:
use push opcode on stringbuilder to fix jumptable code generation
m-eta-wink-kzd: 619f97b | plobsing++ | src/ojs.coda.js:
semi-colon discrepancy fix
m-eta-wink-kzd: ea27760 | plobsing++ | src/Ī©Ī·.capo.winxed:
coerce hash iterator key to string
plobsing fbrito: it gets called when there is an existing value for the key and the key is composite (example: ['composite';'key'], counter-example: ['simple-key']) 06:48
fbrito ahh
plobsing sleeps
fbrito thank you very much
good night :)
cotto fbrito, found one 06:50
oh
dukeleto cotto: did you need something? 06:55
cotto good question. Did I ping you? 06:56
cotto remembers
dukeleto cotto: you mentioned me having admin bits
cotto I think there's a post-commit hook on parrot/pir that needs to be removed
dukeleto cotto: why so? 06:57
cotto it was causing duplicate karma for PIRATE commits
nothing serious
dukeleto cotto: actually, something needs to be deleted from dalek-conf.json from the perl6/mu repo 06:58
cotto dukeleto, why? Isn't it better to have all configuration in one place?
dukeleto cotto: read this github.com/perl6/mu/blob/master/mi...k-push.txt 07:01
cotto: github.com/perl6/mu/blob/master/mi...-conf.json
sorear++ fixed the issue an hour ago :)
cotto: also, please take a look at my new addition about keeping a fork in sync in git_workflow.pod 07:02
cotto I saw.
dukeleto cotto: that should help GCI students, now that there is only 4 days left...
but will be useful for anyone forking on github...
cotto for single-purpose forks master isn't too bad, but I agree that we should recommend that most work go in a branch. 07:03
dukeleto cotto: single-purpose fork? 07:04
cotto: if someone wants to ever sync up to upstream (git rebase), then committing to master leads to pain 07:05
so i just recommend against it.
Anyone smart enough to ignore the advice in git_workflow.pod and still get stuff done is more than welcome to.
I just don't want to answer the same questions any more :)
cotto that's a good motivation
I mean someone who forks Parrot to make a single change and never plans on changing anything again 07:06
but even the user forking can't know for sure if that'll be the case
dukeleto cotto: sure, in that case, committing to master is fine.
cotto: yep. as insurance for the future, it is best to recommend a branch, but if they ignore that and never sync, there is no harm.
i have forked many github repos planning to work on them and then I never make a single commit. 07:07
07:08 theory left
dukeleto cotto: i have one gcc compile farm machine running "make smoke" once an hour on master. It ain't much, but it is a start. 07:08
cotto: if there are specific config flags that we want to test (because perhaps they are uncommonly chosen), let me know 07:09
cotto we should at least have both --optimize and normal
fbrito (wow. I have typed Ā§P0 instead of $P0 and lost 10 minutes trying to figure out why I was getting this error: error:imcc:syntax error, unexpected $endInvalid character in ASCII string)
cotto perhaps c++ too, though that gets some kind of regular testing 07:10
dukeleto fbrito: that's the way she goes, sometimes.
fbrito++ for figuring it out :)
cotto fbrito, did you find a decent test case for that code? 07:11
dukeleto cotto: i am only doing --optimize right now
fbrito cotto which code?
dukeleto fbrito: we are talking about testing parrot with a c++ compiler 07:12
cotto: perhaps some of the NCI Configure.pl flags? Which config flags are most likely to break?
cotto fbrito, the orderedhash code you were trying to cover a bit earlier this evening
fbrito cotto: I am trying to test the compound keys 07:14
cotto yup 07:15
fbrito cotto: And there is no PIR opcode that calls set_*_keyed_str and delete_keyed_str. Removing them (except set_pmc_keyed_str, that is called in internal code) didn't affect tests or build, but I feel that I should research a little bit more 07:18
cotto fbrito, that's a longstanding stupidity of Parrot. 07:21
07:22 KatrinaTheLamia joined, KatrinaTheLamia left
cotto My recommendation would be to either make delete_keyed a dispatcher to deleted_keyed_x (same for the others) or nuke x_keyed_x 07:23
07:23 KatrinaTheLamia joined
fbrito what I have right now is here: github.com/fernandobrito/parrot/co...rderedhash 07:27
it increased the test coverage from 92.7% to 98.4%. I am trying to get it closer to 100%
cotto what's missing 07:31
?
fbrito those if clauses with "next" (plobsing mentioned the composite keys, but they are not working as expected) and something else... wait 07:32
07:34 aantn left
cotto do you have a coverage report you can nopaste or something? 07:36
07:36 fperrad joined
sorear dukeleto: You misread the scrollback. I still need you. 07:37
cotto actually I can generate that myself 07:39
sorear dukeleto: Or rather, I still need the attention of someone with parrot org admin access. 07:40
fbrito cotto: line 349, 674, and 739 of tapir2.ro.vutbr.cz/cover/cover-resu...h-pmc.html 07:42
those are the lines that I haven't covered 07:43
cotto 349 is easy enough 07:45
nopaste "cotto" at 192.168.1.3 pasted "coverage for line 349" (6 lines) at nopaste.snit.ch/27575
dalek rrot/leto/embed_grant: 720c466 | dukeleto++ | t/src/embed.t:
[t] Parrot_PMC_(assign_pmc|add|defined)
 07:48
07:53 aantn joined
cotto the others look similar 07:55
fbrito hm, interesting 08:00
it didn't cover the line 349 for me :o 08:02
cotto odd. I stuck a fprintf in there and it got hit 08:03
wrong file 08:04
nopaste "cotto" at 192.168.1.3 pasted "for real this time" (8 lines) at nopaste.snit.ch/27576 08:05
fbrito I couldn't read those values that were just set 08:08
I am getting "1" instead
aantn asides from "parrot t/testfile.t" is there another way to run one test at a time? 08:09
I want to see an overview of each tast (i.e. PASS or FAIL)
fbrito aantn: you can run them with "prove"
"prove t/..." 08:10
aantn: also take a look on "prove -h"
aantn fbrito: thanks
fbrito cotto: this works: pastie.org/1433375 but if I change from Hash to OrderedHash, it will return 11 instead of "bc" 08:11
cotto that's special. 08:14
Intuitively it looks wrong, but I need to get to bed. If you suspect that it's an incorrect behavior, please file a ticket. 08:15
fbrito I am also going to bed
thank you for your help
cotto OrderedHash used to be more hacky than it is now, so it hasn't gotten as much love as Hash.
kapace__ good night everybody
cotto 'night
08:43 cogno joined 08:59 cogno left 09:04 cogno joined 09:18 cogno left 09:24 cogno joined
dalek a: 02251a6 | fperrad++ | lua/ (2 files):
fix the following error:
 09:32
09:43 ppant left 09:51 rfw left
dalek lscript: abca4e7 | fperrad++ | dynext/pmc/wmlsbytecode.pmc:
string_make was removed of Parrot
 09:54
09:57 ligne joined, ligne left 10:14 bacek joined 10:17 cogno left
dalek p-rx: 412953d | bacek++ | src/setting/IO.pm:
Improve c<slurp> to be more consistent with Perl6 and work in utf8/binary mode
 10:18
p-rx: b812ff7 | bacek++ | src/stage0/nqp-setting.nqp:
Reboostrap setting only.
rrot: d8405fa | bacek++ | ext/nqp-rx/src/stage0/nqp-setting.nqp:
Bring new nqp setting with utf8-aware slurp
 10:19
rrot: e5b39b7 | bacek++ | src/ (3 files):
Merge branch 'master' of github.com:parrot/parrot
tadzik bacek: wanna take a look at my .worry, while near nqp-rx? 10:20
bacek tadzik, nope :) It's slightly out of my responsibility. 10:21
tadzik, I mostly work on nqp setting.
tadzik oh, ok :)
I'll wait for pm then 10:22
bacek tadzik, you can ask sorear also
tadzik sorear: you around? 10:23
it could use some refactoring though 10:24
10:27 cogno joined
sorear tadzik: for a few seconds longer 10:29
10:30 M_o_C joined
tadzik sorear: could you take a look at my patch in the nqp-rx GH issues? 10:31
sorear no 10:32
not now
tadzik ok
10:47 fbrito left 11:10 contingencyplan left 11:45 cogno left, cogno joined
mikehh t/pmc/packfileopmap.t FAILS in make corevm/make coretest - Could not find oplib `math_ops' (passes make test after make world) 11:54
bacek mikehh, oops. packfileopmap shouldn't be in coretest than 12:03
12:10 cogno left 12:23 Drossel left 12:31 JimmyZ joined 12:32 JimmyZ left 12:41 bluescreen left 12:46 kid51 joined
mikehh bacek: maybe you can check for 'math_ops' and skip if not there - see t/pmc/integer.t (check for 'sys_ops') 12:54
12:55 cogno joined 13:04 Kulag joined 13:10 cogno left 13:20 kid51 left 13:34 whiteknight joined
whiteknight good morning, #parrot 13:37
13:43 lucian joined, rurban_ joined 13:45 rurban left, rurban_ is now known as rurban
whiteknight msg bacek PackFileBytecodeSegment.load_lib looks like it enters into an infinite loop 13:51
aloha OK. I'll deliver the message.
13:51 M_o_C left 13:58 kapace_ joined, aantn left
whiteknight msg plobsing can you take a look at TT #749 when you get a chance? A GCI student implemented this but you reverted it. If we really don't want to go this route for now can you close the ticket? Thanks 13:58
aloha OK. I'll deliver the message.
13:59 kapace__ left 14:15 lucian left 14:27 aantn joined, bluescreen joined 14:28 DPini joined, DPini left, DPini joined 14:32 Coke left 14:33 aantn left 14:41 mtk left 14:46 mtk joined 15:07 Patterner left, Psyche^ joined, Psyche^ is now known as Patterner 15:22 Myhrlin left 15:33 DPini left
whiteknight is there anybody here with a working rakudo build? 15:35
whiteknight can't build Rakudo on his little VM 15:36
15:36 fbrito joined
dalek TT #856 closed by whiteknight++: segfault in pbc_merge 15:40
TT #856: trac.parrot.org/parrot/ticket/856
moritz whiteknight: last I tried rakudo built fine on newest parrot, but had about 5 spectest failures 15:41
which was yesterday 15:42
15:44 lucian joined 15:47 contingencyplan joined
whiteknight moritz: I have a rakudo-related ticket that I'm trying to triage. Would you be able to take a look at TT #802? 15:47
it's 18 months old, and I'm trying to figure out if it's still a problem or not 15:48
moritz takes a look 15:50
15:50 donaldh joined
moritz turns out I need to build rakudo first :-) 15:52
dalek TT #1128 closed by whiteknight++: Possible memory error in compilers/imcc/parser_util.c:INS() 15:57
TT #1128: trac.parrot.org/parrot/ticket/1128
16:04 mtk left 16:05 mtk joined 16:08 donaldh left 16:09 lucian left
moritz bah, rakudo on latest parrot is broken again :( 16:13
16:23 [hercynium] joined 16:24 chromatic joined
tadzik aloha 16:25
16:27 [hercynium] is now known as hercynium
moritz oh hai zebra 16:28
tadzik o/ 16:29
dalek rrot: ea9f63e | Whiteknight++ | docs/embed.pod:
start to bring docs/embed.pod up to date with current embedding API
16:29 plobsing left
tadzik did anything funny happen today? :) 16:29
16:29 dmalcolm joined
dalek TT #1591 closed by whiteknight++: PackFile_new_dummy is not in embed.h nor extend.h 16:30
TT #1591: trac.parrot.org/parrot/ticket/1591
whiteknight stupid PackFile_new_dummy
msg plobsing I want to get rid of PackFile_new_dummy. I think this shouldn't be too hard, but there are a few strategies we could follow to do it. I would like your input on it before I start ripping. 16:31
aloha OK. I'll deliver the message.
cotto whiteknight, that code sounds special. What does it do? 16:39
16:43 JimmyZ joined
fbrito should I file a ticket, or is this the appropriate behavior of OrderedHash? pastie.org/1434417 16:46
16:46 JimmyZ left 16:50 plobsing joined
moritz whiteknight: re rakudo related segfault, the PIR example isn't valid syntax anymore (probably realted to opcode removal) 16:50
tt #802 that is 16:51
whiteknight moritz: great. 16:59
moritz whiteknight: I've updated the ticket, loading io_ops helped 17:00
whiteknight cotto: If there is no packfile loaded into the interpreter, attempting to load_bytecode or similar currently fails because load_bytecode attempts to do a merge. PackFile_new_dummy creates a new empty packfile in the interp to allow new things to be merged in
moritz++ 17:01
17:01 gbacon joined 17:02 JimmyZ joined 17:23 plobsing left 17:36 gottreu joined 17:38 contingencyplan left 17:40 contingencyplan joined 17:43 contingencyplan left 17:44 DPini joined 17:45 contingencyplan joined 17:46 Kristaba joined 17:47 gottreu left
Kristaba cotto: ping 17:47
17:49 gottreu joined 17:52 theory joined
dukeleto mornin' 17:55
whiteknight: +1 to killing PackFile_new_dummy
tadzik o/
dukeleto whiteknight: i see you made a huge change to the embed docs 17:58
cotto_work ~~
Kristaba: pong
dukeleto whiteknight: you removed some of the docs that i just added as part of my TPF grant 17:59
Kristaba cotto: Hey, what about my task problem today? 18:00
dukeleto whiteknight: we still want to keep the old docs around while the old embed system is deprecated, for people that are still using the old embed api (like everyone)
whiteknight: with a large DO NOT WRITE NEW CODE WITH THIS API warning
whiteknight: what about docs/embed_old.pod ? And then when the old embed api is deleted from a supported release, we remove embed_old.pod ? 18:01
18:01 theory left
Kristaba cotto: Did you find the time to look at that? :) 18:01
dukeleto whiteknight: for instance, for people that want to hack on converted PL/Parrot to the new embed API, it is useful to compare the old docs and new 18:02
s/converted/converting/
cotto_work Kristaba: I've verified that the branch I'm working with does have the zeroarg changes. 18:08
and your changes 18:09
18:09 theory joined
dukeleto whiteknight: is there no way to choose runcores with the new embed API? 18:10
Kristaba cotto_work: And about the differences between make fulltest and make test? 18:11
cotto_work I will look into those now.
Kristaba ok
dalek rrot: 40e018d | Whiteknight++ | docs/pdds/pdd27_multiple_dispatch.pod:
remove :invocant TODO from the PDD27
 18:14
cotto_work Kristaba: I only see failures in t/compilers/imcc/imcpasm/opt1.t and t/compilers/imcc/imcpasm/opt2.t with fulltest, probably because our fulltest target runs a bunch of tests in serial and dies when one batch fails. 18:17
with test I see many more failures.
dalek rrot/gci_orderedhash_coverage: 7e13bd7 | fbrito++ | src/pmc/orderedhash.pmc:
Remove unused set_*_keyed_str and delete_keyed_str from OrderedHash
 18:18
rrot/gci_orderedhash_coverage: 5c61023 | fbrito++ | t/pmc/orderedhash.t:
[t] Increase OrderedHash test coverage
rrot/gci_orderedhash_coverage: 50f36a3 | fbrito++ | t/pmc/orderedhash.t:
[t] Increase OrderedHash test coverage even more
rrot/gci_orderedhash_coverage: b39915c | fbrito++ | t/pmc/orderedhash.t:
[t] Increase OrderedHash test coverage to 100%
rrot/gci_fix_zeroargs_tests: 81b0439 | (LĆ©o Grange)++ | t/ (12 files):
Fix all tests which failed in zeroargs_subs_checking branch
 18:19
dukeleto fbrito++
Kristaba cotto: Ok, so I'll simply fix all the test failures reported by make test?
dukeleto Simple Matter Of Programming.
cotto_work Kristaba: yes
dalek rrot/gci_tt1199: 984cb27 | Yuki`N++ | / (2 files):
Mod config script to tee output to config.log
 18:20
18:25 JimmyZ left
dalek TT #483 closed by whiteknight++: :invocant flag 18:25
TT #483: trac.parrot.org/parrot/ticket/483
18:28 plobsing joined, darbelo joined
cotto_work seen tcurtis 18:29
aloha tcurtis was last seen in #parrot 1 days 22 hours ago joining the channel.
dukeleto Yuki`N++
karma N
aloha N has karma of 50.
dukeleto giggles
cotto_work That N is really a stand-up guy. 18:30
tadzik (Yuki`N)++ 18:32
karma (Yuki`N)
aloha (Yuki`N) has karma of 0.
tadzik nah
cotto_work karma Yuki`N 18:33
aloha Yuki`N has karma of 0.
dukeleto Evidently, nobody with a ` can have karma. Who knew? 18:47
karma Little`Bobby`Tables";drop database; 18:48
aloha Little`Bobby`Tables";drop database; has karma of 0.
18:49 darbelo left 18:52 davidfetter joined
cotto_work bacek: ping 18:52
moritz $ ./perl6 -e 'say 1' 18:54
===SORRY!===
Confused at line 1, near ".loadlib '"
that's on newest parrot
fbrito dukeleto: thank you for having a look on my task :) 18:56
cotto_work moritz: will try to duplicate 18:57
19:00 davidfetter left
Kristaba Can someone explain me how to debug this kind of error "current instr.: 'anon' pc 0 (EVAL_1:1)"? 19:01
19:01 dzoe left
Kristaba I don't understand what the EVAL_1 "files" refers to 19:03
19:05 bkuhn joined
moritz it means that it's not directly compiled from a file, but has gone through an eval() or so 19:05
dukeleto Kristaba: it basically means you are inside an anonymous function evaluation 19:07
19:08 plobsing left 19:09 DPini left
Kristaba Okay, thanks 19:10
19:16 davidfetter joined
cotto_work moritz: verified 19:17
checking if it's something fulltest catches 19:20
19:29 plobsing joined 19:30 rfw joined
dalek TT #1935 created by pmichaud++: nqp-rx on Parrot RELEASE_2_11_0-771-g40e018d fails t/nqp/33-init.t 19:31
TT #1935: trac.parrot.org/parrot/ticket/1935
rrot: 766fed5 | cotto++ | MANIFEST.generated:
update MANIFEST.generated after opmap_aware_pmc merge
 19:36
19:36 Khisanth left
cotto_work whiteknight: are you looking into the breakage? I want to get this fixed. 19:41
whiteknight yeah, I'm on it
stupid tailcalls are stupid
pmichaud note that it reportedbly breaks rakudo as well.
cotto_work is that what's broken?
pmichaud I'm testing that and will file a ticket on that too.
cotto_work pmichaud: confirmed that rakudo's broken
even hello world
whiteknight cotto_work: I suspect so. The issues are looking very similar to problems I fixed yesterday in #1029
pmichaud so, we don't need a ticket?
whiteknight i'm exploring that issue now 19:42
cotto_work pmichaud: no. There's no way we'll forget about this.
19:42 particle1 left
cotto_work Whatever failure this is isn't caught by fulltest. 19:45
pmichaud according to the nqp failure, it has something to do with :load subroutines 19:46
(rakudo fails its 'make test', yes.)
19:46 particle joined
cotto_work whiteknight: have you tested if rakudo works with that #1029 fix reverted? 19:47
whiteknight cotto_work: working on it
pmichaud (hint: test nqp -- it's faster :)
cotto_work or that
moritz currently bisects the nqp failure 19:48
I'm close to finishing 19:49
last or second-to-last step
pmichaud it looks to me as though the :load subroutines aren't executing in sequence.
(but that's just a guess at this point) 19:50
oh, nm
something else must be happening
cotto_work nqp-rx passes all tests with the 1029 fix reverted (apart from one related to ICU, whcih I haven't installed
whiteknight yeah, it's stupid .tailcall again
cotto_work )
whiteknight damnit
cotto_work If Rakudo hello world works, we need to revert that fix. 19:51
whiteknight I'm working on it
cotto_work awesome
pmichaud I've been trying to eliminate .tailcall from all of my sources
19:51 plobsing left
pmichaud (because it's been problematic with profiling as well) 19:52
moritz my bisect blames 6231646a848d71d90185bbeb0ef4b8383f29a767, which I somehow doubt
cotto_work pmichaud: couldn't you just hack tailcall to behave like a normal call?
dalek rrot: 9328cff | Whiteknight++ | compilers/pct/src/PCT/HLLCompiler.pir:
remove a .tailcall optimization from HLLCompiler.pir.
rrot/zeroargs_subs_checking: 81b0439 | (LĆ©o Grange)++ | t/ (12 files):
Fix all tests which failed in zeroargs_subs_checking branch
whiteknight I can't build rakudo here. So can somebody test that fix with Rakudo? 19:53
pmichaud cotto_work: I don't understand
cotto_work whiteknight: will do
whiteknight ls
cotto_work .
..
moritz no such file or directory 19:54
cotto_work I'm checking how the revert of 8b188db does first
whiteknight: if your fix doesn't work, that one looks promising
19:55 Khisanth joined
whiteknight cotto_work: 8b188db? You have a link? 19:56
cotto_work github.com/parrot/parrot/commit/8b188db
dalek rrot: 4bc0723 | nwellnhof++ | / (2 files):
[io] Remove platform specific code from socket_api.c
rrot: 84f8d95 | nwellnhof++ | / (2 files):
[io] Remove Parrot_io_is_encoding

Unneeded after unicode_io merge
rrot: d5d4070 | nwellnhof++ | / (3 files):
Fix headerizer warnings
whiteknight oh, right. Yes, my most recent commit reverts part of 8b188db 19:57
cotto_work may look familiar
pmichaud fwiw, I'd prefer that the .tailcall in HLLCompiler be simply used to track down a confirmed parrot bug -- not that we want the optimization permanently applied 19:58
there's almost no significant improvement to be had by using the .tailcall
(other than we get failures)
I'll add that note to the ticket
dalek TT #1029 reopened by coke++: taillcall into PIR compiler broken 20:04
TT #1029: trac.parrot.org/parrot/ticket/1029
pmichaud afk, lunch 20:06
Kristaba bacek: ping 20:09
20:12 bluescreen left
cotto_work nqp-rx and Rakudo seem to be back to normal 20:17
whiteknight I can't get over how stupid IMCC is 20:18
NEED_CONTINUATION is defined as (PMC*)0x01
and during all my testing, the PIR compreg always set interp->current_cont = NEED_CONTINUATION
but in some tests, it appears to be setting interp->current_cont = NULL instead
so my previous fix didn't catch that, and NQP explodes
cotto_work two things: 20:20
1 - We need automated HLL testing. This is my new mission in life.
2 - can you add a test so that this particular bit of idiocy doesn't bite us or HLLs again?
whiteknight way ahead of you 20:21
cotto_work that's not too hard
dalek rrot: ef82f5c | Whiteknight++ | / (2 files):
remove special-case code in NCI.invoke. We never assume the interp->current_cont coming out of func is valid now. Add a test to prove that this no longer breaks
whiteknight the test is ugly, but does fail before this fix, and passes after
cotto_work tests can be ugly 20:22
dukeleto: ping 20:23
dukeleto cotto_work: pong 20:28
cotto_work dukeleto: setting up automated HLL testing of the latest Parrot just got important. Does the gcc compile farm allow any sort of web access?
dukeleto cotto_work: that is exactly where i am blocked right now. I don't know that answer to that. I most likely have to ask
cotto_work dukeleto: ok 20:29
dukeleto cotto_work: why did it recently get a lot more important?
dukeleto backlogs a little
cotto_work we broke Rakudo again 20:30
and automated testing most likely would have caught it
dukeleto cotto_work: yes, it would 20:32
cotto_work: i plan to use jitterbug to automate testing (in addition to using plumage)
cotto_work: github.com/franckcuny/jitterbug 20:33
cotto_work: i've spend the last few days improving jitterbug, because i am using it at $work as well
cotto_work: i have a commit bit
cotto_work I see 20:34
dukeleto cotto_work: sgn-build-64.sgn.cornell.edu:8090/ is an example
cotto_work: it integrates with perlbrew to be able to test under all versions of perl that perlbrew knows about
cotto_work: so we can test parrot with all the perls we support
whiteknight okay, my latest fix fixes .tailcall for the nqp failures 20:35
cotto_work dukeleto: that's good, but we really need testing against HLLs.
whiteknight I'm not going to replace it in the HLLCompiler code, but if we did I think it would work
cotto_work whiteknight: I verified that you fixed it.
dukeleto cotto_work: i hear ya
whiteknight I'm going to re-close #1029 now
dukeleto cotto_work: how about you start a wiki page and describe all the features that we need from automated testing? 20:36
pmichaud I'd like to also interject here the comment I just made on #perl6 to the effect that I'm disappointed with the breakages, but I'm equally disappointed with the ongoing attitude of committers who "fix bugs reported by rakudo devs" but don't actually check that the reported bug was really fixed.
whiteknight cotto: there are two "fixes". The first was removing .tailcall. The second was fixing NCI.invoke
I've done both
cotto_work ok
pmichaud whiteknight: I'm not sure that #1029 can be closed.
dukeleto cotto_work: and how important each feature is relative to others
pmichaud the original reported bug apparently still exists.
cotto_work dukeleto: that's a good idea
whiteknight pmichaud: I've put in a fix, added a test for the fix, and verified that NQP continues to work if we chose to replace the .tailcall. What else would be needed to close it?
dukeleto pmichaud: i just want to say that I don't dismiss HLL concerns
pmichaud dukeleto: I did say committers. 20:37
dukeleto pmichaud: but i do agree that Parrot and HLLs currently don't effectively communicate
pmichaud dukeleto: I agree that Parrot's managers (in their manager role) have been extremely response.
*responvie.
*responsive.
20:37 fbrito1 joined
pmichaud whiteknight: oh, I didn't see the ef82f5c fix (although I did look for it) 20:37
nm then -- I agree we can close the ticket, and Whiteknight++ 20:38
afk, kid pickup from school
whiteknight pmichaud++ Thanks for your patience. I can close #1935 too?
pmichaud yes.
dukeleto cotto_work: i don't think smolder does what we need now. It is a black hole of information, has no notifications or summaries, and loses lots of test reports 20:39
I am not happy with Smolder.
20:39 plobsing joined
whiteknight dukeleto: so the million dollar question: What can we get that's better? 20:40
cotto_work dukeleto: I'll think about what's needed, start a wiki page and we can iterate from there
whiteknight do we fix/patch smolder?
dukeleto whiteknight: what do we need?
cotto_work time for food
dukeleto whiteknight: have you read the source of smolder? The code for it is not bad, but it has an amazing amount of dependencies, and there are bugs in the dependencies
whiteknight: like memory leaks and such 20:41
20:41 fbrito left, fbrito1 is now known as fbrito
dukeleto whiteknight: the better question is: What do we actually need? 20:41
whiteknight memory leaks are not necessarily a huge problem, on a short-lived server process
plobsing ~~
whiteknight dukeleto: I have been perfectly happy with smolder. For the most part I think it addresses the needs I have
dukeleto whiteknight: until recently, a smolder server had to be restarted daily or it would consume all memory
whiteknight other community members may be more demanding 20:42
fbrito (can someone please close www.google-melange.com/gci/task/sho...9360922947 ? :))
dukeleto whiteknight: well then you can continue to enjoy to use it :) I don't want to get rid of it, i just don't think it serves all our needs
fbrito: looking
whiteknight dukeleto: I'm certainly not going to argue about anything. I don't love smolder
but I am not inclined to spend my own tuits replacing it
dukeleto fbrito: closed 20:43
whiteknight: i hear ya
Kristaba cotto: I need the confirmation of bacek, but I think all the tests fails are due to the compilers generated code. If it is, I think I'm unable to fix the code in less than few days... 20:44
plobsing whiteknight: (re: TT #749) (1) the commit I reverted did not address this issue. (2) we in fact *do* do some validation of fpmcs before thawing that is shared from pbc thaw. 20:46
I'm not entirely sure what that ticket expects
whiteknight plobsing: okay, if you don't understand it, and I dont uderstand it enough to even know if a fix was committed to it, maybe we just close it 20:47
plobsing: I recently did add a verification function for fpmcs used by the embedding API to verify serialized config hashes. I don't know if I duplicated effort there
plobsing I'd expect cotto to know what was expected. He did report it in the first place.
whiteknight: there is already sanity checking built in to imageiothaw 20:48
whiteknight Parrot_pf_verify_image_string
dukeleto whiteknight: did you see my comments earlier about embed docs? 20:49
plobsing the problem with having the sanity checking in the deserialization (where it belongs) is that it is difficult to handle exceptions generated in a sane fashion.
whiteknight dukeleto: no
dukeleto whiteknight: lost in the sea of irc
whiteknight msg cotto_work when you get back can you take a look at TT #749. I have some questions about what it really wants and whether it's worth doing at this point 20:50
aloha OK. I'll deliver the message.
dukeleto 09:58 <@dukeleto> whiteknight: i see you made a huge change to the embed docs
10:00 <@dukeleto> whiteknight: we still want to keep the old docs around while the old embed system is deprecated, for people that are still using the old embed api (like everyone)
10:01 <@dukeleto> whiteknight: what about docs/embed_old.pod ? And then when the old embed api is deleted from a supported release, we remove embed_old.pod ?
10:02 <@dukeleto> whiteknight: for instance, for people that want to hack on converted PL/Parrot to the new embed API, it is useful to compare the old docs and new
10:10 <@dukeleto> whiteknight: is there no way to choose runcores with the new embed API?
dukeleto should have used a link, but was lazy
plobsing so we "need" a separate validator so that we can make decent error messages (eg: for fpmc config hash thaw in fakecutables)
whiteknight: the *real* problem is that we don't have a decent (or at least well known) mechanism for catching thrown exceptions from within C internals 20:51
dukeleto whiteknight: also, some of the work that i have done for my current TPF grant was removed in your embed.pod commit
whiteknight dukeleto: Ah, I wasn't aware that it conflicted with your grant 20:52
I'll bring back the old file and rename it to embed_old.pod
dukeleto whiteknight: danke
whiteknight and include a note that the newer file is for the newer API
dukeleto whiteknight: awesome. I can add a mention in embed_old to the current one 20:53
dalek TT #1029 closed by whiteknight++: taillcall into PIR compiler broken 20:54
TT #1029: trac.parrot.org/parrot/ticket/1029
TT #1935 closed by whiteknight++: nqp-rx on Parrot RELEASE_2_11_0-771-g40e018d fails t/nqp/33-init.t
TT #1935: trac.parrot.org/parrot/ticket/1935
plobsing whiteknight: I feel such separate validation functions are a workaround for a completely separate issue. They are hackish, wrong, and if I break them, I will simply remove them.
whiteknight plobsing: That function is basically just an encapsulation for preexisting logic that was in an even worse place 20:56
so if you want to remove it and you don't think it's worthwhile, do it
I didn't create the logic, I just moved it to a function in the correct subsystem
it had been in src/global_setup.c, I believe
plobsing whiteknight: yes. it was added to give a "nice" error message for fakecutables linked against the wrong libparrot and close out a ticket.
the correct fix would be difficult, so this happened in stead. 20:57
whiteknight plobsing: what is the correct fix?
plobsing catch the thrown exception
whiteknight the embed API does that, in a fashion
plobsing append to the error message to give more context
moritz whiteknight: is there test coverage for trac.parrot.org/parrot/ticket/1935 in the parrot test suite? 20:58
whiteknight moritz: Not directly, but yes.
I added a test for it and #1029
so the tailcall issue there should be resolved
20:59 bacek left
plobsing I have no immediate desire to eliminate it. There are bigger fish to fry. But I have no intention of expending effort to keep it working. 20:59
moritz whiteknight: ah, the test is hidden in ef82f5ccda42b44738c2eca023693fd95d10b092 21:00
well, not hidden
pmichaud does this help, at all? gist.github.com/768575 21:01
moritz just at the end of teh line :-)
whiteknight moritz: yes
cotto_work ~~
whiteknight pmichaud: yes
plobsing whiteknight: (re: PackFile_new_dummy) we can get rid of this as soon as we no longer special-case the first packfile. 21:02
whiteknight plobsing: Okay, awesome. That's what I want to do 21:03
cotto_work whiteknight: what about 749?
whiteknight I want to get rid of that, and I want to remove some of the special-case crap from the embedding API. I want one function for "Add a packfile to the interpreteR" 21:04
pmichaud should I post the nqp-rx steps to parrot-dev, or ... ?
whiteknight instead of one function to load the first one, one to load subsequent ones, one to run it, etc
pmichaud: parrot-dev
plobsing one function to load them function to load them all, one function to run them... 21:05
whiteknight cotto_work: Do we need that ticket still? with other refactors in that system the ticket may be redundant
dukeleto pmichaud: steps to test nqp-rx + parrot latest together? 21:06
cotto_work whiteknight: closed\\ 21:07
whiteknight cotto_work: we'll make it better en passant, I promise 21:08
cotto_work yes we will
pmichaud message sent to parrot-dev 21:09
21:10 bacek joined
whiteknight I was very happy to see opmap_aware_pmcs merged 21:10
bacek++ times infinity for that one
dalek TT #749 closed by cotto++: [CAGE] factor Packfile_unpack into validation and unpacking functions 21:11
TT #749: trac.parrot.org/parrot/ticket/749
whiteknight we're down to 575 tickets 21:12
cotto_work indeed
whiteknight It's possible (unlikely) that we can get down to 550 by 3.0
dukeleto pmichaud++ # concise instructions 21:13
pmichaud: i can setup something to do that on the GCC compile farm pretty easily
pmichaud: are there any Configure.pl flags for the parrot that is used that you prefer ? 21:14
pmichaud I tend to use --optimize
and of course a --prefix for installing
so, --optimize and --prefix 21:15
Kristaba bacek: ping
whiteknight msg kid51 during your spaces_in_paths work, can you take a gander at TT #196? I'm on MSVC+ActivePerl and I can't reproduce it. When you merge that branch and all my tests come out clean I'm inclined to close that ticket
aloha OK. I'll deliver the message.
dukeleto pmichaud: does nqp-rx have a place to send smoke reports ?
21:16 plobsing left 21:17 fbrito left
whiteknight we can easily add it to smolder if desired 21:18
though nqp-rx would need to write a client to upload the reports
cotto_work one issue that the automation should mitigate is ephemeral failures in nqp-rx and other HLLs under test 21:19
false positives are not going to be helpful
pmichaud dukeleto: no smoke reporting on nqp-rx at the moment. we haven't need them -- test failures are pretty rare 21:20
Kapace will trac.parrot.org/parrot/ticket/1038 be made into a GCI task? It looks interesting
pmichaud if we want to put together some smoke reporting for nqp-rx, I be supportive of that 21:21
afk, homework
*I'd
cotto_work dukeleto: do you have tuits to quickly turn that into a gci task?
Kapace: start on it if you're interested. I'll make it a task tonight if nobody else does. Just send a pull request when you're ready. 21:25
Kapace cotto_work: ok cool 21:26
21:28 plobsing joined 21:29 whiteknight left
bacek aloha, humans 21:37
cotto, Kristaba pong
cotto_work bacek, can you change aloha to allow backticks in nicks? 21:38
dukeleto cotto_work: i have no time right now to do anything 21:40
cotto_work dukeleto: ok
dukeleto cotto_work: smolder.parrot.org/app/projects/smoke_reports/8 21:42
nqp-rx has a home on smolder now
should be trivial to send reports to it
cotto_work: feel free to make it a GCI task
21:43 gottreu left 21:44 rurban_ joined, Coke joined
dukeleto pmichaud: smolder.parrot.org/app/projects/smoke_reports/8 <-- nqp-rx is on smolder now 21:44
21:46 rurban left, rurban_ is now known as rurban 21:47 mtk left
Kristaba bacek: For my fix task, I fixed all the test I knew fix (all tests that failed using make fulltest except in t/compilers/imcc/opt1.t and opt2.t), but now, with 'make test', I have a lot of test failings 21:49
cotto_work Kristaba: do you have a parrot/parrot commit bit or did someone else pull your commits? 21:51
Kristaba bacek: And I think they are dues to the compilers code output (so I really have no idea of how to fix them)
cotto: Someone else pull my commits 21:52
nopaste "Kristaba" at 192.168.1.3 pasted ""make test" error output with my changes" (619 lines) at nopaste.snit.ch/27578
Kristaba bacek: See the nopaste if you have some time 21:53
bacek: I think the HLL compiler need to be fixed, but I have no more idea of the cause of this issue 21:56
bacek Kristaba, pct test failures are bad. They have to fixed (either tests or PCT by it self). You can ignore IMCC failures.
Kristaba, last test - just change OFFSET=>13 to 15 21:57
cotto_work Kristaba: you're being extraordinarily patient with a task that was probably too difficult for gci. 21:59
Once you're done with this round of changes, I'd be ok considering your part complete. Do you want me to mark it complete so you can move on?
bacek PCT failures - :main subs should have ".param args :slurpy" 22:00
(It's from #1705)
Kristaba bacek: Oh, thank you! 22:01
22:01 cognominal left
Kristaba cotto: Thank you too, I'm ok for this :) 22:02
cotto_work excellent tenacity, then 22:03
22:04 aantn joined
cotto_work dukeleto: are you happy with the config log gci task? 22:04
dalek r: ea8a091 | bacek++ | src/POST/Compiler.pm:
Simplify emiting of POST::Label
 22:08
r: 902650c | bacek++ | src/POST/Compiler.pm:
Fix emiting bytecode for method call.
r: 10601f1 | bacek++ | src/POST/Compiler.pm:
Charset is gone.

Handling of strings require major rehaul.
22:10 nwellnhof joined 22:19 aantn left 22:21 Matt221 joined
Matt221 Hi guys 22:21
sorear Hello
Matt221 Just FYI: I wont be able to finish the 'Take 100 lines off fill_params' GCI task until tomorrow (swamped with school work)
22:22 hercynium left
cotto_work Matt221: fbrito is also taking a related task. Please coordinate with him so you're not both changing the same code. 22:23
Matt221: never mind. He's not on it any longer. 22:24
22:25 cognominal joined
cotto_work Matt221: I look forward to seeing what you come up with. 22:25
dukeleto cotto_work: yes, i think it is better than what we have 22:30
cotto_work dukeleto: so approve?
dukeleto cotto_work: yes 22:31
cotto_work done
dukeleto cotto_work: you making a task to add "make smoke" to nqp-rx ? 22:32
cotto_work: the code can basically be copy-pasted from rakudo
cotto_work i wasn't planning on it
wouldn't be too hard though
dukeleto cotto_work: writing the task will take longer than just doing it
dukeleto has an nqp-rx commit bit
cotto_work It's getting close to the point where I don't see much value in adding tasks. 22:33
apart from ones that we know for certain a student will work on
dukeleto cotto_work: yep
Kapace How's this looking: dpaste.com/295039/ 22:36
cotto_work ... 22:38
speedy 22:39
Kapace (I had been looking at that ticket from earlier :P) 22:40
cotto_work does that expose everything that the current version does?
Kapace Only has 3 exposed subroutines, so yeah, as far as I can tell
i don't think the internal functions are of much use... 22:41
cotto_work no
make sure to add some tests and docs 22:42
task created 22:43
Kapace ah, yes, forgot about that. I remade the example for comparison, but I'll make some tests
ah, thanks cotto_work
cotto_work sure. if you're trolling through our tickets and find anything else before the gci deadline, let me know 22:44
Kapace cool
cotto_work do you know the parrot-specific parts of adding a file to our repo? 22:45
Kapace no
but, I'm not adding a source file, just a test maybe..
cotto_work ok'
that's easy then
Kapace (there are no existing tests for md5sum, so I probably have to add one) 22:46
cotto_work t/library/md5.t
rfw [11:23:11] <@cotto_work> Matt221: fbrito is also taking a related task. Please coordinate with him so you're not both changing the same code. <-- no actually this is me
but i dropped it to take up a sahana bug 22:47
cotto_work rfw: ok. I unpublished it for now.
rfw alright
i'll pick it up when Matt221 finishes
cotto_work if you want to do it, let me know and coordinate with matt
er, Matt221
rfw actually
can we swap? 22:48
i can do it before Matt221
then he can pick up the task after i'm done
cotto_work Matt221: is that ok with you?
dukeleto tools/bisect-parrot.pl in rakudo.git looks useful
cotto_work rfw: I republished. Go for it, but do your best to coordinate. 22:49
rfw okay
but i have to hang the washing first
cotto_work: claimed 22:51
22:52 nwellnhof left
cotto_work accpeted 22:54
22:55 plobsing left
dalek p-rx/smoke: e1f6b8b | dukeleto++ | build/Makefile.in:
Begin to add a smoke target to send smolder reports
 22:57
Kapace does method documentation go at the header of the file, or before each method? (or can someone point me to a properly documented source file?) 22:59
23:00 plobsing joined
mikehh make corevm/make coretest FAIL - t/pmc/packfileopmap.t - Could not find oplib `math_ops' - remaining tests pass 23:04
all other tests PASS (pre/post-config, smoke (#2098) fulltest) at 2_11_0-777-gef82f5c - Ubuntu 10.10 i386 (g++-4.5)
cotto_work Kapace: before each method 23:06
23:07 whiteknight joined
Kapace ok 23:09
23:10 plobsing left 23:17 fperrad left
Kapace btw, I had to add :main to the current md5 example for it to work 23:22
cotto_work Kapace: ok 23:26
that's not terribly surprising
23:28 bkuhn left
Kapace cotto_work: ok. I see sha256 in Digest too, how about an object interface for that as well, for consistancy? 23:29
cotto_work sure 23:30
claim accepted
Kapace cotto_work: cool 100th pull :) github.com/parrot/parrot/pull/100 23:31
cotto_work I was hoping we'd get there before the end of gci
rfw how does headerizer work? 23:32
it doesn't seem to, uh, headerize, as it were 23:34
cotto_work rfw: you run make headerizer after a build 23:38
rfw oh, okay
cotto_work rfw: what are you doign
rfw cotto_work: ripping fill_params into bits 23:39
since matt is doing it tomorrow i'll do it today :D
cotto_work great 23:40
chose the function names carefully
rfw it won't headerize my new function though
cotto_work do you have a branch?
rfw i can make one for my fork
Kristaba My god, the PCT tests errors make me crazy 23:42
I still don't understand from where the " 'anon' pc 0 (EVAL_1:1) " error can be fixed 23:43
23:50 Matt221 left
Kapace hmm, the output of a simple sha256 test is not equal to one from an online calculator.. 23:58
(the online test concurs with my sha256sum tool, but not the pir program) 23:59

Powered by Cro and the RakuĀ® Programming Language.

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

Thank you!