Parrot 1.6.0 "half-pie" released! | Test CallSignature PMC | pcc_reapply branch still needs your love! trac.parrot.org/parrot/wiki/Callin...nsOverview
Set by moderator on 12 October 2009.
Whiteknight urg. Github isn't allowing me to look at TCL's files 00:01
jonathan has probably created more PMCs that override find_method than most. :-) 00:04
You can do a lot of cool stuff that way. :-) 00:05
Whiteknight yeah 00:06
how do I get a copy of a repo from github?
specifically, I'm trying to pull a local copy of partcl/partcl 00:07
darbelo git clone?
purl git clone is probably at 87%
darbelo purl forget git clone
purl darbelo: I forgot git clone
Whiteknight I tried "git clone git@github.com:partcl/partcl.git" and it said permission was denied 00:08
jonathan git clone git://github.com/partcl/partcl.git
darbelo try git clone git://github.com/partcl/partcl.git
jonathan (the other one is if you want to authenticate, iirc) 00:09
Whiteknight fatal: Unable to look up github.com (port partcl) (Servname not supported for ai_socktype)
darbelo wtf?
Whiteknight nevermind, tried again, worked
jonathan Computers are SO not deterministic some days. 00:10
:-)
darbelo Whiteknight: git@ urls are for repos you have commit bits in.
japhb ENOCHROMATIC, dangit 00:17
OK, I'll open this up to the peanut gallery: Is there a community convention for where to put test data that is more than a single file in size? I was thinking /test-data/, just to make it self-documenting, barring a better suggestion 00:19
(This is for plumage, not parrot)
darbelo I did t/data for decnum-dynopmcs 00:20
japhb darbelo, as long as the test harness knows that directory is special ... plus it kinda makes you think there is a pile of tests for "data" in there, rather than a group of data for other tests ... 00:21
but I'm not strongly against it, it just feels a little off.
It does have the benefit of keeping all testing-related bits in the same tree 00:22
darbelo I was actually storing the data I used to generate my tests. Not acutal test data. 00:23
japhb Hmmmm, perhaps if all test scripts were in t/\\d+-\\w+/ , then t/[anything else] is freed up conceptually. 00:24
dukeleto, thoughts?
darbelo Also, my harness only runst ".t" files in the t/ directory and leaves the subdirs alone. 00:25
Whiteknight is pcc_reapply supposed to be completely broken right now? 00:37
darbelo define "completely"
But I'd guess no. 00:38
00:41 rhr joined 00:42 rdice joined
japhb dukeleto, darbelo: For git at least, we don't need to ship a test repo -- we can just init one as needed. For subversion, this might be a little harder, unless we can assume svnadmin ships with svn everywhere. 00:44
(It does on my Debian box, FWIW ... but Debian packages tend to be "less broken" than average.) 00:46
Tene Whiteknight: no
Whiteknight: allison merged chromatic's callsignature.pmc rewrite 00:47
Whiteknight: that's the biggest change recently
darbelo japhb: OpenBSD ships the in the same package. Works for me , but I don't really know if it it's the right thing. 00:48
Tene Whiteknight: realcleaned? can you paste the failure?
japhb darbelo, OK, let's go crazy and assume proper packaging until some moon OS proves otherwise. ;-) 00:49
darbelo fwiw I've found the BSD's tend to not break up upstream tarballs into a gazzillion sub-packages, like some linux distros do. 00:51
00:51 davidfetter_ joined 00:55 darbelo left
japhb darbelo: Proper, that. I find the "gazillion packages that you somehow end up downloading all of anyway" to be infuriating. On Debian, OpenOffice is unfortunately packaged like this. 00:56
Whiteknight Tene: nevermind, I think I had some local mods 00:57
works just swell
Tene Whiteknight: Okay. :)
00:59 rhr joined 01:00 kid51 joined
japhb dukeleto, a couple new things in Plumage's TASKS, just making above ideas concrete. 01:05
japhb now thinking about dep handling
01:06 Eevee joined
dalek rrot-plumage: 33b0f44 | japhb++ | :
[META] New TASKS, and update an old one
 01:07
shorten dalek's url is at xrl.us/bfsf6j
Whiteknight I don't see anything in thread.c that is obviously relying on old-style argument passing 01:11
01:11 rhr joined 01:19 rhr joined
Whiteknight the thread argument PMC is coming out as garbage inside the NCI invoke 01:20
what is "PP@" as a signature string? 01:22
I've never seen an @ in a sig before
(and I suspect it is an error) 01:23
solved one failure 01:28
dalek rrot: r41867 | whiteknight++ | branches/pcc_reapply/tools/build/nativecall.pl:
[pcc] fix several (but not all) failures in threads.t. Problem was a bad signature string being automatically generated with '@' instead of 'Ps' for a slurpy array
 01:29
japhb dukeleto, See latest parrot-plumage commit for more fun stuff to work on. :-) 01:38
purl, msg darbelo See latest parrot-plumage commit for more fun stuff to work on. :-)
purl Message for darbelo stored.
japhb dalek will notice the new commit about the time I'm eating my pizza, I think. :-) 01:39
Whiteknight I don't see a "join" method defined anywhere in ParrotThread
dalek rrot-plumage: 9d32829 | japhb++ | :
[META] Two more (big) groups of TASKS
shorten dalek's url is at xrl.us/bfsga6
01:42 mikehh joined 01:45 rdice joined
Whiteknight I might have a fix for more tests coming up too 01:53
we're down to 9 failing tests now 01:56
that's single digits, baby!
dalek rrot: r41868 | whiteknight++ | branches/pcc_reapply/tools/build/nativecall.pl:
[pcc] invocants should have a signature of Pi in NCI calls, not just P. This doesn't affect any tests, but still is a change that needs to be made
 01:58
02:05 rhr joined
Whiteknight that's all I have for now. Goodnight 02:05
mikehh pcc_reapply - make smoke #29021 at r41867 - 10,752 ok, 13 failed, 262 todo, 560 skipped and 0 unexpectedly succeeded - Ubuntu 9.10 beta amd64 02:09
02:28 rhr joined 02:34 JimmyZ joined 02:41 janus joined
dalek rrot: r41869 | mikehh++ | branches/pcc_reapply/src/pmc/callsignature.pmc:
fix codetest failures - unwrapped macro arguments and space between function name and parenthesis
 02:47
02:55 JimmyZ joined
dalek rrot: r41870 | mikehh++ | branches/pcc_reapply/src/call/args.c:
fix codetest failure - space after opening parenthesis
 02:57
rrot-plumage: 24db387 | leto++ | :
[configure] Mention how to run the test suite when Configure says hello
 03:49
shorten dalek's url is at xrl.us/bfsgu9
dalek rrot-plumage: 7d43fa6 | leto++ | :
Merge remote branch 'origin/master'
shorten dalek's url is at xrl.us/bfsgvd
dukeleto yuck, my merge commit got pushed 04:31
a mock git repo, consisting only of one commit with an empty .gitignore file is only 92K. after I delete the sample post-commit hooks, 42K 04:38
1609 bytes compressed. not bad at all 04:42
msg japhb how should i compress the mock repo? currently I have it as a .tar.gz . but what happens if tar/gzip are not available? 04:45
purl Message for japhb stored.
japhb Did you see my comment about using 'git init' and 'svnadmin create' during test setup instead of shipping a mock repo? 04:46
dukeleto, feel highlighted 04:47
dukeleto japhb: oooh, i didn't realize that you meant doing that at test setup time. sneaky!
japhb :-) 04:48
dukeleto japhb: so do we assume that git and svn are on the system?
japhb: or should we detect both and not run those tests if one/other is not present?
japhb we can't test the fetch without them ... my feeling is, we check if they are there, and ... yeah, what you just said
dukeleto japhb: do we actually check for svn/git being installed currently? 04:49
japhb dukeleto, that's part of the latest TASKS
;-)
dukeleto i like the idea of creating at test time 04:50
japhb: i read TASKS, but perhaps too quickly :)
japhb heh
well, a couple relevant items were ...
1) make a function that acts like the 'which' shell command (path search)
2) Use %VM<config> to find the local executable name for each of the basic build programs, and use the function from 1) to detect them 04:51
BTW, if one of the programs we want doesn't have a %VM<config> entry, for the time being lets just default to assuming it is the unix name plus the platform executable extension (e.g. .exe) 04:52
dukeleto, make sense to you? 04:53
dukeleto japhb: i think that I will write tests that check for a value in %VM<config> for svn/git binaries, if it doesn't find any, then we just go with the default that you described 04:54
japhb: yes, mostly
japhb yep, that's what I meant
Oh, I see,
you meant do 2) without 1) 04:55
Fair enough.
Baby steps.
purl baby steps! baby steps!
dukeleto japhb: so 'which' iterates over entries in PATH, opens each directory and searches for the binary in each
japhb inch-pebbles
dukeleto, yup.
dukeleto japhb: yes, 2) without 1) in anticipation of 1) getting done :)
japhb heh
dukeleto japhb: i can make the tests check if the initial git/svn command returns an error and skip the rest of the tests 04:56
japhb: i am in need of a function like qx() which gives me the return code of the command as well
japhb Grrr, parrot doesn't record the platform PATH separator. I guess we have to start writing a platform hints section ... 04:57
dukeleto japhb: yak stacks within yak stacks
japhb: currently I would have to run every test twice in the plumage test suite to get the return code and the output
japhb: so I just don't check return codes currently 04:58
japhb dukeleto, I'm not sure parrot currently provides what you want out of qx() ... and yeah, I saw where you were going with the question. ;-)
There are a number of places in the IO subsystem of Parrot that are ... clearly prototypes at best.
And your idea for just paying attention to whether git init succeeds is a good one 04:59
dukeleto japhb: yes, i will just check the status code on running git/svn the first time, and skip the test file if they don't exist 05:00
japhb nodnod
dukeleto looks like dalek doesn't notice commits to branches other than master 05:02
japhb Yeah, I think right now that's the only Atom feed dalek is watching. 05:03
I'm not sure what the other feed URIs would actually be, come to think of it 05:04
dukeleto it's not a bid deal, just noticing
japhb: is it using gitorious.org/parrot-plumage.atom ?
japhb: that should include all branches 05:05
japhb Hmmm, I think it's gitorious.org/parrot-plumage/parrot-plumage.atom , IIRC
japhb chuckles at mode changes in last commit 05:07
dalek rrot-plumage: d3ff89f | leto++ | :
[cage] Fix permissions in t/
shorten dalek's url is at xrl.us/bfsg29
dalek rrot-plumage: 95caad6 | japhb++ | :
[META] TODO updates
shorten dalek's url is at xrl.us/bfsg3b
dalek rrot-plumage: 05dc282 | japhb++ | :
Merge branch 'master' of git@gitorious.org:parrot-plumage/parrot-plumage
shorten dalek's url is at xrl.us/bfsg3d
japhb Oy vey those merges are annoyin
g
dukeleto japhb: yes, things were exactly opposite what they should have been
japhb: sorry about that. i need to git pull --rebase
japhb no, me too 05:08
05:08 JimmyZ joined
japhb what's the config key to make that default again? 05:08
dukeleto japhb: oh yeah, that is *your* mere commit
but I have a few as well
japhb: there is one, but I forget
japhb found it
dukeleto japhb: enlighten us? 05:09
japhb Sorry, was actually just typing it in myself. :-)
'git config branch.<branch>.rebase true' 05:10
dukeleto japhb: thanks 05:11
japhb: we need a way to mock out the repo URL in the metadata. is it possible to provide plumage with an alternate metadata directory? 05:12
japhb checking ...
purl i guess checking ... is it complicated to get to work with google and yahoo ?
japhb not right now, but give me a sec, I might be able to hack it in. 05:13
dukeleto shweet 05:15
japhb stick that in your pipe and smoke it. 05:18
You just have to provide a config file that overrides plumage_metadata_dir now.
Or at least it *should* work. :-) 05:19
dukeleto japhb: nice! that should all me to start hacking on the tests 05:22
japhb excellent
dukeleto japhb: does t/metadata/ sound good to you?
japhb: next question: can local paths be used in the metadata file? 05:23
dalek rrot-plumage: 449259e | japhb++ | :
[plumage] Allow plumage metadata dir to be overridden by conf file
shorten dalek's url is at xrl.us/bfsg36
dukeleto japhb: uck. I meant relative paths
japhb dukeleto, yes, t/metadata/ is fine 05:24
relative paths to what?
dukeleto japhb: the repo URLs. i don't know that I need that or if it even makes sense 05:25
dukeleto is still fighting a cold and is a bit out of it today 05:26
japhb The repo urls just have to be understandable by the tools, not by us.
dukeleto japhb: yeah. i need to fiddle with repository{checkout_uri} and make it a local file. 05:27
japhb: i think i can use file:// in there and git and svn should know what to do
japhb dukeleto, cool beans 05:28
dukeleto japhb: what does Plumage::Downloader get used for? deps?
japhb It will be used sometime in the future for several things. The two most important are: fetching projects that are in tarballs instead of source repos (especially useful for fetching release versions :-), and for interacting with the metadata server 05:30
dalek p-rx: c5a7893 | pmichaud++ | (4 files):
Add Regex::P6Grammar as temporary scaffolding for bootstrapping the parser.
 05:31
shorten dalek's url is at xrl.us/bfsg49
dalek p-rx: 2bb5b72 | pmichaud++ | .gitignore:
Update .gitignore.
shorten dalek's url is at xrl.us/bfsg5b
dalek p-rx: f80ba14 | pmichaud++ | (2 files):
Change "make test" target to specify $(PARROT) executable (suggested
dukeleto japhb: ok, so when we need to fetch tarballs, that is when we need to make P::Downloader understand file:///, to be properly tests. just getting things straight in my head
purl dalek: that doesn't look right
shorten dalek's url is at xrl.us/bfsg5d
japhb dukeleto, exactly so 05:32
dukeleto s/tests/tested/, but you got the gist
shweet
05:49 uniejo joined
japhb dukeleto, those last couple commits might make your life easier 05:52
dalek lag ...
dalek rrot-plumage: f5d478f | japhb++ | :
[CORE] Glue.pir: Add path_exists()
 05:56
shorten dalek's url is at xrl.us/bfsg6p
dalek rrot-plumage: fd0f294 | japhb++ | :
[plumage] Switch to path_exists() everywhere
shorten dalek's url is at xrl.us/bfsg6r
dalek rrot-plumage: ec431fc | japhb++ | :
[plumage] Add find_program() utility function
shorten dalek's url is at xrl.us/bfsg6t
dalek p-rx: e7e5924 | pmichaud++ | src/Regex/P6 (2 files):
Add the ability to parse arglists after colons in subrule calls.

Bootstrap progress: The "p6grammar" executable is now able to parse its own grammar definition; test via
   ./p6grammar --target=parse src/Regex/P6Grammar/Grammar.pm		 06:11
shorten dalek's url is at xrl.us/bfsg78
dalek p-rx: e04a163 | pmichaud++ | src/Regex/P6Regex/Grammar.pm:
Add parsing of simple single-quoted literals in regexes.
 06:22
shorten dalek's url is at xrl.us/bfsg8x
dalek p-rx: 41bdc6b | pmichaud++ | src/Regex/P6Regex/Actions.pm:
Add quoted literals to regexes.
shorten dalek's url is at xrl.us/bfsg8z
dalek p-rx: 0dfed1f | pmichaud++ | STATUS:
Update STATUS a bit.
shorten dalek's url is at xrl.us/bfsg83
dalek p-rx: 75fd363 | pmichaud++ | src/Regex/P6Grammar/Grammar.pm:
Use single quotes instead of backslashes for quoted literal punctuation.
p-rx: 3b052a5 | pmichaud++ | src/Regex/P6Regex/Grammar.pm:
Clean up a few regex parsing definitions in P6Regex.

   ./p6grammar --target=parse src/Regex/P6Grammar/Grammar.pm
   ./p6grammar --target=parse src/Regex/P6Regex/Grammar.pm
shorten dalek's url is at xrl.us/bfsg85
shorten dalek's url is at xrl.us/bfsg87
Tene pmichaud: excellent work
japhb pmichaud++ 06:24
pmichaud thanks
next step will be to have P6Grammar generate a standalone parser that parses its own
grammars. 06:25
purl grammars are classes, and classes are dynamically modifiable... hmmmm....
pmichaud then we won't need Perl6Grammar anymore :)
or PGE, for that matter. 06:26
(others will need it, yes, nqp-rx won't need it)
japhb Excellent. 06:27
How is the progress towards replacing the old NQP (to a level that I could switch Plumage over)?
pmichaud well, this is all part of that. 06:28
What is currently p6grammar in the nqp-rx repo is the beginnings of the new NQP.
japhb Poorly worded, sorry. I meant, what is the current estimated time frame?
pmichaud I don't know for sure. I'd hope just a few days. Somewhat depends on schedule and timing around the house.
getting to this point has been quite a bit easier than I had anticipated. 06:29
(it's taken longer than I anticipated, but that's due to lots of other factors)
however, I'm finding that each new feature I add is easier to do than the ones that came before it 06:30
06:33 JimmyZ joined
japhb That's got to be a great feeling .... 06:33
g'night all!
06:49 fperrad joined 06:53 barney joined 07:01 chromatic joined 07:18 bacek joined 07:19 s1n joined
chromatic The good news is that we've cut off 31.183%. The bad news is that we still have a ways to go. 07:26
If we double the speed in the branch, we come out ahead.
Doubling that speed won't be easy though.
bacek o hai 07:29
chromatic hey 07:30
bacek chromatic: how is going with CallSignature hacking? 07:33
chromatic We merged it into pcc_reapply today. 07:34
mikehh we are down to 13 failing tests 07:40
dalek rrot: r41871 | pmichaud++ | trunk (4 files):
[nqp]: Add simple class inheritance to NQP.
 07:42
chromatic I have 9.
mikehh Well that's that's what smolder reports #29022 at r41870 07:44
chromatic Maybe my count is off. 07:45
bacek chromatic: why you reserving 3 bits? Not 2? 07:46
mikehh t/library/coroutine.t - Parse errors: Bad plan. You planned 6 tests but ran 2 07:47
probably accounts for the difference - (4 tests missed - but counted as failures by smolder) 07:48
chromatic bacek, I should be reserving two.
If I have the math wrong, I won't act surprised.
Tene bacek: he masks against ~3 because 3 is 1+2
so just 2 bits 07:49
bacek ah.. I misread code...
chromatic We could turn that magic number into an intelligible #define though.
... and fix the comment. 07:50
dalek p-rx: bbbca64 | pmichaud++ | src/Regex/P6Regex/Actions.pm:
Refactor named assertions slightly for better argument handling later.
 07:56
shorten dalek's url is at xrl.us/bfshf8
dalek p-rx: 2b0f1ad | pmichaud++ | (3 files):
More updates to P6Grammar bootstrapping.

for its own grammar definition (in src/Regex/P6Grammar/Grammar.pm). Still need a metachar:<{*}> action method to produce action methods. :-)
shorten dalek's url is at xrl.us/bfshga
08:05 iblechbot joined 08:18 AndyA joined
dalek rrot: r41872 | mikehh++ | trunk/compilers/nqp/t/30-subclass.t:
set svn properties
 08:25
08:25 FullMetalHarlot joined
dalek rrot: r41873 | mikehh++ | trunk/compilers/nqp/src/Grammar/Actions.pir:
fix codetest failure - trailing whitespace
 08:32
08:39 mokurai left
bacek chromatic: (your patch in list) Main problem with :flat args 08:50
We have to build proper signature in run-time based on passed arguments 08:51
chromatic I wasn't sure. One of the tests that it failed didn't have flat args.
bacek chromatic: around line 557. 08:53
dissect_aggregate_arg returns new signature bit
chromatic Oh, somehow I overlooked flat_list. 08:54
We could make that function manipulate the RIA instead. 08:55
If you feel motivated to replace all of that STRING fun, feel free. 08:58
The good news is that we have another 43% faster to make the branch to reach the trunk's speed on fib.pir. 09:01
bacek I don't think that RIA will be faster...
dalek rrot: r41874 | bacek++ | branches/pcc_reapply/src/call/args.c:
[uncage] Poke more into CallSignature guts for performance sake.
rrot: r41875 | bacek++ | branches/pcc_reapply/src/call/args.c:
Create final result in single shot, not repeatedely in loop in dissect_aggregate_arg.
bacek chromatic: can you benchmark with my latest commits? 09:04
chromatic Running now.
dalek rrot: r41876 | bacek++ | branches/pcc_reapply/src/pmc/callsignature.pmc:
Reorder checks in CallSignature.get_attr_str to put "named" on top. We are mostly use this attribute.
rrot: r41877 | chromatic++ | branches/pcc_reapply/src/call (2 files):
[PCC] Removed unnecessary GC registration and unregistration of CallSignature

unnecessary (and on platforms where it is necessary, we'd have other GC problems). This provides a 6.36% performance improvement in the fib.pir benchmark.
purl It seems faster
rrot: r41878 | chromatic++ | branches/pcc_reapply/src/call/args.c:
[PCC] Removed unnecessary (and too frequent) MMD type tuple creation in

it itself if necessary.
rrot: r41879 | chromatic++ | branches/pcc_reapply/src/pmc/callsignature.pmc:
[PMC] Fixed comment about tag bits in CallSignature and turned a magic number
chromatic Hah, we have overlapping commits.
Three in the last five minutes.
No change, bacek.
bacek strange... 09:06
purl well, strange is but true
bacek I expected 0.1% improvement
chromatic Big gains: merging CallSignature and Context, removing use of CPointer. 09:07
We might be able to cut out 2/3 of our PMC allocations if we do that. 09:08
Huge win.
Modest improvement: preallocate appropriate size of returns array in Parrot_pcc_build_sig_object_returns_from_op. 09:10
bacek What about :slurpy? 09:11
chromatic No idea. 09:12
bacek btw, op set_returns still have gc_unregister_pmc 09:13
chromatic It goes! 09:14
bacek chromatic: my benchmark says 16,013,422,581 vs 15,199,915,860 on fib.pir at rr41873 41876 (unoptimised build) 09:18
chromatic I have an optimized build. 09:19
5.08% though, not bad. 09:20
bacek We probably have to put PARROT_CALLING_START/_END inside call_from_sigobject 09:25
to properly mark C stack 09:26
but I'm not sure about it.
chromatic I'm not sure either.
bacek 14,330,100,550 on most recent commit 09:27
09:28 donaldh joined
bacek purl: (16013422581-14330100550)/16013422581 09:28
purl 0.10511944104924
chromatic Handy.
bacek chromatic: it was your commits :) 09:29
chromatic A couple more and I'm calling it a night. 09:31
43.2% to go. 09:35
dalek rrot: r41880 | chromatic++ | branches/pcc_reapply/src/ops/core.ops:
[ops] Removed another gc_unregister_pmc() now unused from set_returns op. See
 09:37
p-rx: 468b4eb | pmichaud++ | src/ (3 files):
Add keyed reductions / action triggers.
 09:40
shorten dalek's url is at xrl.us/bfshr3
dalek p-rx: 0c574ea | pmichaud++ | src/ (2 files):
Determine regex name to be used for action methods.
shorten dalek's url is at xrl.us/bfshr5
dalek p-rx: 672674d | pmichaud++ | src/Regex/P6Grammar/ (2 files):
Update P6Grammar to attach names to regexes, set modifiers for
shorten dalek's url is at xrl.us/bfshr9
09:41 masak joined 09:43 payload joined 09:44 payload joined
bacek purl: (14330100550-14255582083)/14330100550 09:56
purl 0.00520013566827345
bacek yet another half percent :)
dalek rrot: r41881 | bacek++ | branches/pcc_reapply/src/call/args.c:
Preallocate string_sig in Parrot_pcc_build_sig_object_from_op
 09:59
10:20 integral joined 10:39 JimmyZ_ joined, cconstantine joined 11:01 plobsing joined, allison joined 11:33 jan joined 11:38 einstein joined
Coke msg dukeleto pinged perl webmasters about the planet feed. 11:55
purl Message for dukeleto stored.
12:04 whiteknight joined
whiteknight good morning parrot 12:05
12:28 mikehh joined
mikehh pcc_reapply - make smoke #29042 at r41881 - 10,752 ok, 13 failed, 262 todo, 560 skipped and 0 unexpectedly succeeded - Ubuntu 9.10 beta amd64 12:33
12:35 bluescreen joined
whiteknight 13 failed? We we down to 9 last night before I logged off 12:39
at least, on my system 12:41
some good optimization work happened last night though, so that's good to see 12:43
12:59 davidfetter joined
Coke I need to get my own server to run builds on. :| 13:03
13:03 payload joined
moritz Coke: I can give you access to a quad core machine with linux amd64 13:04
Coke moritz: did you already do that?
moritz not that I know of
Coke Someone gave me access to a box that I was unable to get parrot to build on, and then I got distracted. =-)
moritz I gave japhb access to it for his plumage meta data server 13:05
and I had no troubles building parrot on that machine
Coke azawawi?
moritz no
Coke no, that's who's eating up the CPU on feather atm. =-) 13:06
moritz he has no access on that other machine :-)
Coke moritz++ 13:22
moritz :-)
look, your build is finished :-) 13:23
13:30 mikehh joined
nopaste "mikehh" at 81.149.189.7 pasted "pcc_reapply branch - summary of fulltest failures at r41881" (47 lines) at nopaste.snit.ch/18348 13:32
mikehh whiteknight - it is true that 9 failed but 4 were missed because t/library/coroutine.t exits after test 2 - Parse errors: Bad plan. You planned 6 tests but ran 2 13:36
and you get Failed 4/6 subtests which is picked up by smoke/smolder 13:37
have a look at the TAP output in smolder.plusthree.com/app/public_pr...ails/29042 13:40
shorten mikehh's url is at xrl.us/bfsic4
mikehh or if you like what shorten suggests 13:42
Coke mortiz++ # again 13:43
moritz karma mortiz 13:44
purl mortiz has karma of 2
13:50 whiteknight joined
Coke hah! 13:56
mortiz--
moritz++
14:03 whiteknight_ joined
dukeleto 'ello 14:09
14:10 Psyche^ joined
dalek kudo: d5a2ee9 | moritz++ | CREDITS:
restore alphabetical order in CREDITS
 14:13
shorten dalek's url is at xrl.us/bfsihp
14:19 schmalbe joined
dalek rrot: r41882 | fperrad++ | trunk/tools/dev/fetch_languages.pl:
[languages] Partcl moves on Github
 14:21
14:22 whiteknight_ joined
whiteknight internet connection today is TEH SUXXOR 14:26
14:35 theory joined
dalek rrot: r41883 | fperrad++ | trunk/tools/dev/fetch_languages.pl:
[languages] matrixy moves on Github
 14:37
14:42 JimmyZ joined, Limbic_Region joined 14:44 Hunger joined
Coke Limbic_Region: hey 14:46
dukeleto funny URL of the day: www.nonblocking.io/ 14:49
moritz wanted to register std.io but was too slow
Coke moritz: spectest is, oddly, passing more tests on your box. =-) 14:50
I was just doing a speed comparison against feather, but this is odd.
Limbic_Region salutations Coke - love to stay and chat but I am TTFN 14:53
15:03 schmalbe joined
Coke moritz: ooh, segfault. :| 15:10
15:11 hercynium joined
moritz innocent 15:11
purl well, innocent is guilty...
moritz Coke: did you also test on amd64 before?
Coke likely not 15:12
15:12 allison joined
Coke moritz++ 15:18
Tene hi allison
allison hi Tene 15:19
Tene: just got back from classes 15:20
Tene: looks like quite a few checkins in pcc_reapply
Tene: (she says, hoping for more passing tests)
moritz less, actually - because tests that we previously never reached are run now, and fail 15:21
wait
still more passing tests
purl more passing tests are always useful
allison moritz: just also more failing tests
moritz right
dalek p-rx: ec4637c | pmichaud++ | (6 files):
Change src/parrot/ to more correctly named src/cheats/ .
 15:24
purl dalek: that doesn't look right
shorten dalek's url is at xrl.us/bfsisa
allison moritz: still, down from 18 failing tests to 12 failing tests, that's progress 15:32
whiteknight I was down to 9 failing tests when I turned off the computer last night 15:34
allison msg chromatic I wonder if it's worth merging CPointer down to an internal array in CallSignature instead of a separate PMC? We're planning to rip it out in January anyway, but the speed gains might be worth it. 15:35
purl Message for chromatic stored.
whiteknight fixed some issues in the nativecall.pl
Tene allison: he's already planning to do that, I believe. 15:36
allison Tene: good 15:39
dalek rtcl: d1c5cc3 | (Will Coleda)++ | docs/spectest- (2 files):
Spec test run for amd64;
 15:40
shorten dalek's url is at xrl.us/bfsiua
Tene allison: at least, he's already mentioned "getting rid of CPointer" a few times as part of his goals here. 15:44
allison Tene: yes, we're definitely getting rid of it in January 15:45
Tene: but, we hadn't talked so far about getting rid of it now
Tene: (but, a lot happens during the time I'm asleep in this timezone)
Tene That is, "stop using cpointer for results passing" 15:46
allison Tene: yes 15:47
Tene: in january we'll just be doing results passing the same as arg/param passing 15:48
Tene Right. 15:50
whiteknight getting rid of CPointer in the CallSignature logic would hugely reduce GC pressure, I think 16:02
and for essentially zero extra complexity. The space for the PMC* to the CPointer could be reused to hold another void* pointer to the arg
mikehh allison: I don't know if you saw, but I put up a summary fulltest report in nopaste.snit.ch/18348 16:13
whiteknight okay, that corroborates the numbers I saw last night. 9 failures in coretest 16:16
t/library/coroutine.t looks like the next big source of errors to target
allison mikehh: thanks, that's helpful 16:27
whiteknight: so, the 'resume' method in runtime/parrot/library/Parrot/Coroutine.pir is taking a parameter of 'args' as a :slurpy, and returning that same variable as a :flat 16:30
whiteknight so are we not flattening returns? 16:31
allison whiteknight: and yet, it's complaining about :flat on a non-array
whiteknight: we are flattening returns 16:32
whiteknight: perhaps it's an off-by-one error in the flattening...
whiteknight ok
allison whiteknight: see src/call/args.c starting around lin 1725 16:33
line
whiteknight okay, let me update 16:34
ok, so that's the error that's being thrown? 16:36
I'm going to try to rebuild, but last time I did on win32 it broke horribly
allison whiteknight: the error is "flattened return on a non-array" 16:38
whiteknight no, same problem
(doesn't build on win32)
that raises two questions: (1) are we sure the return in question is an aggregate and (2) is this the behavior we really want in this case? 16:39
it seems to me that :flat on a non-aggregate would just be that PMC itself
allison whiteknight: digging down in gdb, it's getting a :flat flag on an Integer PMC 16:41
whiteknight the resume method is tailcalling the resume continuation 16:42
allison which would seem like the argument after the :optional
(the :opt_flag) 16:43
dalek p-rx: fbb15ef | pmichaud++ | src/Regex/Cursor.pir:
Add lang argument to Cursor.!cursor_start.
shorten dalek's url is at xrl.us/bfsi5y
dalek p-rx: fb80e2d | pmichaud++ | src/Regex/P6Grammar (2 files):
[p6grammar]: Action methods to automatically create grammar object.
shorten dalek's url is at xrl.us/bfsi52
dalek p-rx: 8cd492a | pmichaud++ | src/Regex/P6Grammar/Actions.pm:
[p6grammar]: Add ability to declare a parent grammar.
shorten dalek's url is at xrl.us/bfsi54
cotto_work did chromatic's branch ever get deleted? 16:45
dukeleto cotto_work: which branch? 16:46
purl hmmm... which branch is this? bacek's?
dukeleto purl, forget which branch
purl dukeleto: I forgot which branch
allison cotto: pcc_optimize_sig? 16:47
cotto_work the one where he reimplemented callsignature
that one
purl it has been said that that one is silly to. It's an impl detail, right?
allison cotto: not yet
16:49 theory joined 16:51 darbelo joined
dukeleto what happened to docs.parrot.org/parrot/latest/html/...e.pod.html ? it used to be a valid link and is now a 404 17:01
shorten dukeleto's url is at xrl.us/bfsjat
dukeleto shorten: you should check if something is a 404 before shortening it. just sayin'
whiteknight we can't make the bots too intelligent! 17:02
dukeleto: pages in the book all got moved around
you'll have to look through the repo /docs/book/* to find it's current resting place
(thanks moritz++)
moritz don't we have bots for opping? 17:04
opbots: trust whiteknight
slavorgn But I already trust whiteknight
slavorg But I already trust whiteknight
moritz I see that.
whiteknight yay! I'm trustworthy apparently
dukeleto looks like it moved to docs/book/draft/
moritz whiteknight: no, you are trusted ;-)
purl okay, moritz.
whiteknight no, you? 17:05
whiteknight: no you?
purl good you.
moritz whiteknight?
purl whiteknight is mailto:wknight8111@gmail.com or the grand master funk or wknight8111.blogspot.com/
whiteknight shakes his fist at purl
17:07 Austin joined
dukeleto looks like the docs/book/draft directory does not get html-ified on docs.parrot.org, shouldn't it be? 17:08
it was publicly accessible and people linked to it, now it got moved around and is not even accessible, which sucks.
whiteknight I dont know what black magic puts those docs on the web in the first place, but yes, I think it should be there
moritz I think it's Coke's black magic 17:10
Austin Are there really still different cores? 17:13
whiteknight cores? 17:14
Austin And if so, does executing a 7 trace 1 switch them out?
run cores
whiteknight oh yes, there are different runcores
I am not familiar with "trace 1"
Austin Trace is an opcode.
1 would be an argument to the opcode.
whiteknight yes, I just don't know how it does what it does
Austin Ah.
Well, the --help-debug suggests that running with --trace uses the slow core. 17:15
But it doesn't say what happens if I turn on trace via opcode.
(What happens is that I get a segfault.)
Not at the site of the trace1, but later, trying to return from a sub call. 17:16
whiteknight maybe that's what is supposed to happen
moritz then we should rename it to segfault ;-)
Austin Indeed.
Put it right next to "clone"
whiteknight we still need to fix that clone nonsense I guess 17:17
whiteknight has been reminded
Austin You mean the other segfault opcode? 17:18
:)
I thought you had that taken care of.
whiteknight I had something that was sort of a solution at one point 17:26
and then it was suggested that both my solution sucked and that I should use it to solve other problems
darbelo whiteknight: ping 17:31
whiteknight darbelo: pong
darbelo I'm looking at matrixy right now, and it's Configure.pl has *ancient* assumptions in it. 17:32
How would you feel about replacing it with something else? 17:33
s/Configure.pl/Configure.pl and Makefile templates/ 17:34
dalek TT #1112 created by brianwisti++: Subject: Draft Parrot Book chapters no longer available on docs.parrot.org 17:35
TT #1113 created by coke++: segfault in ?? 17:38
Coke (docs) no black magic. just "make html" 17:40
dalek p-rx: 07b5fef | pmichaud++ | src/Regex/P6 (2 files):
Add node (line+file) annotations to P6Regex generation.
 17:41
shorten dalek's url is at xrl.us/bfsjik
darbelo whiteknight: Also, what' up with the patch in the top-level dir? 17:43
whiteknight darbelo: the configure script is ANCIENT'
literally, we made that back before Parrot was even installable
so please, if you know how to replace it with something better, please do 17:44
the patch was supposed to add Complex support to PGE, which I think has already been done and can be deleted
there is a LOT of cleanup to do, obviously 17:45
darbelo Hmm. I think I can make Configure.pl work with some minor modifications, gimme a sec. 17:46
17:46 joeri joined
darbelo Hm. No, I can't. Nevermind. 17:48
Coke partcl (unsurprisingly) fails against pcc_reapply branch. 17:49
whiteknight yeah, it's in bad shape
Coke the fake-cutable segfaults. 17:50
whiteknight what runcore does the fakecutable use? 17:51
isn't it the fast core?
17:58 bacek joined
Austin Argh. 18:06
Of course the solution to my segfault problem is to remove a file from the makefile, rebuild, then add it back, re-rebuild. No more segfault. 18:07
Austin is being shagged by a rare parrot.
Coke bad makefile rules.
Coke guesses.
purl guesses are fine, right now.. I don't have a clue what could be wrong :P
Austin www.youtube.com/watch?v=9T1vfsHYiKY
Nah, the makefile rules are fine - it just cats a bunch of stuff into a gen_library file. 18:09
There's something twitchy in the parrot innards that I was bothering. 18:10
whiteknight that's umpossible!
Coke is up to 3 segfaults... 18:11
18:12 chromatic joined 18:15 hercynium joined 18:19 gaz joined
Coke smolder.plusthree.com/app/public_pr...ails/29049 - look at the failure in t/cmd_expr.t ... huh? 18:22
shorten Coke's url is at xrl.us/bfsjq6
Coke (same with the other one.) 18:23
... got "foo", expected "foo" ?
wtf?
kthakore hi 18:27
purl bonjour, kthakore.
kthakore morning chromatic
chromatic morning 18:29
Coke whee, different segfaults for partcl based on amd64 vs. feather. 18:30
Tene hi chromatic.
kthakore: It's nice to see you sticking around. I'm looking forward to seeing patches from you eventually. ;)
dukeleto kthakore: howdy 18:31
kthakore Tene: ok if I see patches for SDL Perl from you ;) 18:32
dukeleto: hit
oops
hi
purl hi, kthakore.
kthakore Tene: I am just trying to get a head start on Parrot for SDL Parrot in Jan
dukeleto kthakore: what are you hackin on in parrot these days?
kthakore dukeleto: line above 18:33
purl line above is the precise line on wich I get the error
dukeleto kthakore: where does SDL Parrot live?
kthakore dukeleto: NCI bindings of SDL Parrot but really really far from now
Tene There are already some SDL bindings...
dunno what state they're in.
kthakore Tene: I know
Tene: teh are the suck 18:34
18:34 AndyA joined
dukeleto kthakore: i am learning about Parrot NCI now as well, since I want to have access to the GNU Scientific Library (GSL) from Parrot 18:34
kthakore dukeleto: ooh any linkies 18:36
dukeleto: SDL is a very modular C library and I am already having trouble with it in p5 but chromatic tells me its eaiser for Parrot 18:37
chromatic I found it was.
kthakore dukeleto: eventually I would like to have one NCI for both Perl 5 and Perl6/Parrot SDL Bindings
Tene "one NCI"? 18:38
kthakore chromatic: so do I
Tene: bindings are in XS right now
Tene Right.
kthakore Tene: P5NCI also exists so if I have NCI then hopefully have Parrot and Perl5 supported from one set of bindings 18:39
search.cpan.org/~chromatic/P5NCI-0....b/P5NCI.pm
Coke planet parrot unstuck, should reflect soon. 18:40
... now, actually. enjoy. 18:41
kthakore Tene: so ya ... one set of bindings. Don't even know if its possible with current P5NCI. 18:43
dalek TT #1012 closed by coke++: Add leto.net/parrot.xml to Planet Parrot
18:44 davidfetter joined
kthakore work time!! 18:44
dukeleto: Tene I be back
Tene bye 18:47
treed Perl question: is there already a mock TCP server? 18:50
like the mock DBI, but for straight TCP
dalek rtcl: 658dde1 | coke++ | (3 files):
Move list of skippable tests into repo instead of pulling from the wiki.
shorten dalek's url is at xrl.us/bfsjwf
treed dammit 18:51
wrong channel
Well, kinda.
Not a totally wrong channel.
But, not my intended channel. 18:52
davidfetter "this is my channel. there are many like it, but this one is mine..."
19:02 AndyA joined
chromatic Moving CPointer into CallSignature would be easy if we could differentiate between a CallSig that handed params and a CallSig that handled returns. 19:04
The returns-filling code creates a new RPA, stores that in the CallSig by attribute, and pushes into the RPA directly. 19:05
dukeleto chromatic: does that imply that that we should perhaps mitosize CallSignature into 2 PMCs?
chromatic Good news: that's two extra PMCs (the RPA and the CPointer) that can go away in the one-parameter case.
I'm thinking about that, dukeleto. 19:06
They store two different things. The param CallSig stores INSPs. The returns CallSig stores pointers to INSPs. 19:07
19:08 desertm4x joined
Coke would it make sense to make them both have pointers or both have not? 19:09
chromatic It's less intrusive right now to retain pointers in the returns case. 19:10
Coke k
chromatic 40% fewer PMCs if I can pull this off though. 19:11
Coke I imagine that would be very nice for partcl.
dukeleto that sounds like quite a performance increase 19:13
dalek p-rx: 40101f8 | pmichaud++ | src/PAST/Compiler-Regex.pir:
Optimize quantification of common character classes like \\s, \\d, \\w, etc.
shorten dalek's url is at xrl.us/bfsj22
chromatic I wonder how long this needs to live, given a pending unification of Context and CallSignature though.
bacek Good morning 19:14
chromatic: It will be useful even after unification
19:15 whiteknight joined
PerlJam chromatic: I don't know if it gets said enough, but once more can't hurt: I really appreciate your efforts to optimize things. It gives me a little thrill to see parrot go faster. 19:16
whiteknight ditto! chromatic++ 19:17
Tene Yes, me too. chromatic++
japhb piles on with a chromatic++ as well
dukeleto chromatic++ # for being lord of optimization 19:18
darbelo chromatic++ # aol
allison chromatic: after the post 2.0 unification, we will have two separate CallSigs, one for args and one for returns 19:24
chromatic: (or, we may recycle the PMC for efficiency, since we don't need the call sig any more when we prepare the return sig)
19:25 mokurai joined
allison chromatic: but, yes, for now we need returns stored inside the call sig, and returns are pointers to INSP instead of INSP directly 19:25
dalek tracwiki: v101 | darbelo++ | Languages 19:29
tracwiki: Point the Matrixy entry to the language's new home in github.
tracwiki: trac.parrot.org/parrot/wiki/Langua...ction=diff
shorten dalek's url is at xrl.us/bfsj5n
allison chromatic: we could store a return CallSignature object inside the call CallSignature object as self->results (instead of a ResizablePMCArray) 19:30
chromatic: and use the unionval trick to store the pointer directly as the positional element of the return CallSignature
darbelo Hmm, we should create a Wiki page for parrot libraries, and teach dalek to scrape it for repos. 19:34
dalek rrot: r41884 | bacek++ | trunk/t/compilers/imcc/syn (2 files):
[t][cage] Fix tests for macro line numbering. Not all cores track
 19:36
TT #1102 closed by bacek++: test failures with t/compilers/imcc/syn/macro.t and regressions.t in fast, ...
chromatic I'm thinking of making another (though I guess if the return size is known, it can be an array) linked list of tagged returns and storing that with get_pointer and fetching it with set_pointer. 19:37
It'd be an array of void *. 19:38
That's super easy to walk, it can be type safe, and it's much, much more compact and GC friendly than another PMC for the container and a PMC for each pointer. 19:39
bacek chromatic: we can precalculate args/returns size. 19:40
Tene exactly what I was thinking of. 19:42
chromatic Wow, look at all the loops in fill_results.
bacek chromatic: or we can explicitly extend internal storage on handling :flar/:slurpy
:flat 19:43
Tene fill_* handles flat and slurpy. no need to deal with them when building the callsig 19:45
and yes, ETOOMANYLOOPS
chromatic The problem is that it's a 507-line C function. 19:46
Tene YES.
chromatic Let's call that Step Zero.
Tene I was planning to look to see if I could unify/factor out anything common between fill_*
chromatic It's like a regular step but has fewer calories and Coke can have a dozen a day.
Unification would be nice, but even extracting individual steps into their own named encapsulation units will help. 19:47
Coke ^_- 19:54
whiteknight so long as the optimization doesn't hurt the ongoing failure debugging, I'm all for it 19:55
some of these failures are rediculous though
bacek chromatic: any objections on storing pointer to last cell in CallSignature to avoid loop in APPEND_CELL? 19:58
Coke msg dukeleto adding a link to the babysteps article you mentioned on teh blog at www.parrot.org would be nifty. 20:02
purl Message for dukeleto stored.
dukeleto Coke: i am confused. where should I add the link? 20:09
20:10 Limbic_Region joined
Coke I was going to suggest just a mini story that pointed to his docs. 20:14
kthakore Tene: back ... 20:18
Tene hi kthakore
kthakore Tene: were we talking about something?
Tene kthakore: No, we weren't.
kthakore ok 20:20
dukeleto Coke: oh, you mean adding a blog post to parrot.org about "baby steps", right ? I was misunderstanding 20:21
Coke yes
dukeleto: regarding missing chapters of books, perhaps: 20:23
"lib/Parrot/Docs/Section/Parrot.pm"
I don't see the text "draft" in that file at all.
dukeleto Coke: yes, it seems that the parrot book got reorganized, but the "make html" target didn't get updated 20:24
cotto_work iirc make html could use a refactor. A couple people have tried to figure it out and given up.
dukeleto cotto_work: yeah, like me :)
cotto_work: i tried to get the dynops docs htmlified, and I got close, but punted 20:25
20:25 bacek joined
dukeleto cotto_work: probably another 30min of fiddling and I will get it to work, but I haven't touched it in weeks 20:25
is there a "goto" person for docs/book/ ? particle? 20:26
cotto_work You could add it to ItsABughunt
I don't know
dukeleto cotto_work: good idea
purl dukeleto: Good Idea: Throwing a surprise party for your father. Bad Idea: Throwing a surprise party for your grandfather.
dukeleto botsnack
purl :)
chromatic bacek, I thought about it, but I'm not sure it adds a lot. It's a minor optimization.
cotto_work It'd save on future party expenses.
bacek chromatic: yeah, but still optimisation. 20:27
hercynium blog post: "FUCK THE CLOUD, BEFORE THE CLOUD FUCKS YOU." ascii.textfiles.com/archives/2255
sorry!
wrong channel :(
chromatic I'm not sure it's a measurable optimization at this point.
dukeleto hercynium: but I am still checking it out :) 20:28
hercynium heh... I'm not a cloud fan m'self
Coke dukeleto: which book?
hercynium (but I *am* a parrot fan)
Coke allison, chromatic, any one else. (roughly in that order.)
20:29 frodwith left
Coke whiteknight, IIRC. 20:29
(if you mean content issues)
dukeleto Coke: any book 20:31
purl rumour has it any book is out of date the moment it hits paper
dukeleto hercynium: yeah, there is a high correlation between people who actively use the term "the cloud" and asshats 20:32
hercynium s/use/sell/
dukeleto hercynium: indeed. 20:34
allison Coke? 20:50
purl i think Coke is Will Coleda <mailto:will@coleda.com> or perpetually annoyed. or magical ticket robot or (if Diet), something that turns into formaldehyde in my blood.
allison oh, goto person for docs/book, right 20:51
dukeleto allison: i was wondering what the deal with the docs/book/draft directory is 20:52
allison dukeleto: yes, chromatic or I have the ability to push changes to update the printed version
dukeleto allison: people linked to ch11 of the the book but then it got reorged and now everything in the docs/book/draft directory is not htmlified 20:53
allison dukeleto: it's the unfinished chapters (either from the old book or new chapters) that didn't make it into the first Parrot book on PIR
dukeleto allison: should they still be publicly accessible as html on the docs.parrot.org site?
allison dukeleto: the drafts are mainly unreadable and unreliable
whiteknight they could be improved
hell, they should be
dukeleto allison: i can understand that, but people linked to them in blog posts and now they are 404s. that sucks
allison whiteknight: sure 20:54
Coke allison: yes?
ah.
dukeleto allison: should I try to fix "make html" to make it include the draft directory?
Coke dukeleto: yes.
(I could have answered that. =-)
allison dukeleto: we aren't going to perpetually preserve all the links in the docs directory
dukeleto Coke: i know you are on board, I just wanted to hear what allison has to say :) 20:55
Coke allison: but it's not included /at all/, even in a draft/ dir.
allison dukeleto: if they want a permenant link, they should link to the page for a particular version of parrot
dukeleto allison: that solves the problem at hand, thanks!
Coke allison: we don't really advertise that link, only latest.
s/advertise/link to, even./ 20:56
allison Coke: the chapters that are reasonably finished and useful in the site are linked in
Coke I would treat the book in that respect like our placeholder code. make available, but mark as experimental.
allison I'd avoid linking in draft chapters that are inaccurate or nonsensical
(which means reviewing them) 20:57
Coke I would rather just publish everything and concentrate on reviewing them.
allison last I checked, some of them don't even parse
if thery're not on the front page, and clearly marked as drafts, it shouldn't throw anyone too far off 20:58
dukeleto just to give a concrete example, someone linked to docs.parrot.org/parrot/1.3.0/html/d...e.pod.html in a blog post, which looks like a useful chapter
shorten dukeleto's url is at xrl.us/bfskox
dukeleto i just changed the link to 1.3.0, but the person was using the link with 'latest' instead of 1.3.0, so the link went dead
Coke IWBNI if .../docs/book/ worked, btw. 20:59
(even if only to redirect back to the top level page.)
dukeleto IWBNI?
purl IWBNI is "it would be nice if"
dukeleto botsnack
purl thanks dukeleto :)
allison dukeleto: except it's wrong in several places
Coke I think we're more like to get our docs fixed if we publish them warts and all (with DRAFT as appropriate). 21:00
allison Coke: a page could be added for that
dukeleto allison: what about putting "THIS IS A DRAFT AND VERY POSSIBLY IS WRONG/OUT OF DATE" at the top of draft chapters?
Coke allison: yes, I'm feeding dukeleto tasks. =-)
allison Coke: the current html doc processing is a bit byzantine
dukeleto coke++ # sunshine is a pretty decent disinfectant
Coke dukeleto: I think "DRAFT" is fine, yes.
perhaps a "NEEDS REVIEW", even. 21:01
allison dukeleto: yes, DRAFT in the title is fine
(or some other large/prominent text
but, not on the front page 21:02
dukeleto allison: so, are you saying that draft chapters should be available but linked to? 21:03
s/linked to/not linked to/
allison dukeleto: how about just one link on the front page to Book? 21:04
dukeleto: then chapter links on the Book page
dukeleto: (or Books)
dukeleto allison: that sounds fine with me.
allison dukeleto: then break that page down into "Programming Parrot: PIR", and "DRAFT", and whatever the other books in docs/book are 21:05
dukeleto: IIRC, there's a rough draft for compiler tools too
dukeleto allison: ok, sounds like a plan
Coke the chapter links for books will include links to draft chapters, yes? 21:11
21:33 cconstantine joined
cotto_work chromatic, do you think the pcc refactor will yield any potential optimizations for the profiling runcore? 22:01
chromatic Nothing comes to mind. 22:05
I had an idea (with Whiteknight? bacek?) a while back about changing the way we index annotations in the annotation chunk of the bytecode.
It's O(n) to find the annotations for any particular opcode, and that adds up for the profiling runcore. 22:06
cotto_work quite
Maybe I should dust off my todo list and get hacking on that. 22:07
chromatic I think we need to fix line numbering first, but japhb had some thoughts on that. 22:08
cotto_work yeah, but that involves imcc 22:11
chromatic That's not the awful part of IMCC. 22:15
Mostly I just need to know which syntactic constructs are wrong. It's reasonably easy to fix then.
japhb chromatic: sorry I have not had the time to deal with that. RL has been in the way last week and this. :-( 22:17
22:22 theory joined
treed Is there a way to get a diff of just the recent commits *I* made? 22:22
(git)
Tene git log | grep treed | ... 22:23
moritz git log --author=treed
and if you also want the diff, add the -p option 22:24
treed Aha, thanks.
Tene Huh. Nice.
moritz++
cotto_work chromatic, do you think we'll eventually want parrot to load libraries that can take CLI options?
Tene "libraries that can take CLI options"?? 22:25
chromatic IMCC versus PIRC, for example? Assuming a PIR parser is loadable.
Suppose the profiling core is loadable but has optional CLI options.
Tene ah
cotto_work yeah, like that 22:26
chromatic cotto_work, eventually yes.
cotto_work ok
22:29 mikehh joined
dalek tracwiki: v5 | cotto++ | ItsABughunt 22:37
tracwiki: add cli argument parsing
tracwiki: trac.parrot.org/parrot/wiki/ItsABu...ction=diff
shorten dalek's url is at xrl.us/bfsk8i
cotto_work Hmm. Processing seems to be the word I was looking for. 22:38
22:55 Whiteknight joined 23:06 patspam joined 23:09 ash_ joined
darbelo Okay, let's see if dalek handles the new repo... 23:26
Whiteknight what new repo? 23:27
darbelo Matrixy. 23:28
Whiteknight oh nice
darbelo I adjusted the matrixy entry on the wiki to point to github. I think that's enough for dalek to start reporting commits.
Whiteknight here's hoping 23:29
japhb darbelo, did you already put in the one for the libs?
Whiteknight there are lots of commits needed :)
darbelo japhb: We don't have a libs page on the wiki.
We should have, and dalek should know about it. But we don't, AFAIK. 23:30
japhb darbelo: I thought we did ... only a couple entries in it, IIRC
darbelo I could be wrong.
japhb looks ....
Ah, here it is: trac.parrot.org/parrot/wiki/Modules 23:31
darbelo And I was the last person to edit it. 23:32
darbelo feels stupuid.
darbelo feels doubly stupid.
japhb heh
No worries. Not like any of us have more than a couple things on our mind at any given time or anything .... 23:33
Whiteknight it's too hard to find pages in that damn wiki 23:35
we need to be better about linking pages together
darbelo Whiteknight: What would you put on the "Written in" column for parrot-linear-algebra? 23:36
Whiteknight PIR
it will probably be a combination of C and PIR
because we're going to need library bindings and wrappers, but we're also going to want matrix PMC types written in C 23:37
darbelo ok, done. Gave it a "Alpha" Completeness. 23:38
Whiteknight nice 23:39
dalek tracwiki: v3 | darbelo++ | Modules 23:40
tracwiki: Add parrot-linear-algebra to the list.
tracwiki: trac.parrot.org/parrot/wiki/Module...ction=diff
shorten dalek's url is at xrl.us/bfsmhy
23:41 theory joined
darbelo But I am pretty sure that dalek doesn't know about that page. 23:41
Whiteknight allison: ping 23:42
japhb darbelo: and thus, your next task. ;-)
Whiteknight allison: nvermind, unping 23:43
I think tailcalls are borked 23:45
chromatic I'm sure they are. 23:46
Whiteknight somewhere the call signature of a tailcall needs to get at the returns portion of the parent's callsignature
and I don't think thta's happening
darbelo trac.parrot.org/parrot/wiki/Languages needs pruning. Lots of abandoned languages there. 23:48
dukeleto darbelo: how do we define 'abandoned' ? 23:49
Whiteknight nobdoy working on it. No hope. All alone and nobody cares.
darbelo Damm, there was a languages/tap in parrot. 23:50
japhb: Hasn't been updated to the brave, new, post-1.0 world. And nobody plans to. 23:51
Whiteknight I think we could save some complexity if set_args returns the CallSignature PMC, and we pass that to get_results 23:52
and then pass it also to invoke
saves us from having to call Parrot_pcc_[get|set]_signature so many times 23:53
dukeleto Whiteknight: 'nobody' working on it for how long?
Whiteknight AND gives us the flexibility to be more manual in our handling of CallSignatures
chromatic Whiteknight, if we unify CallSig and Context, won't we have to undo that? 23:54
dukeleto Whiteknight: just trying to quantify what we consider 'abandoned'
Whiteknight chromatic: no, not at all. We'll still call invoke with the Context PMC
japhb darbelo: huh what?
Whiteknight but that gives us the ability to say "run in this context" a
and gives us the ability to preinitialize a context for a call if we choose
chromatic Which invoke? 23:55
Whiteknight invokecc is the one I am thinking of specifically
darbelo japhb: sorry, meant dukeleto.
Whiteknight dukeleto: I don't know. Anything that doesn't build against current Parrot is a good candidate
anything that hasn't been touched since before the last supported release (1.4) and doesn't even build against that is better 23:56
darbelo dukeleto: I'd call anything that needs a pre-1.0 parrot 'abandoned'
Whiteknight darbelo: I agree
UNLESS it's under active development and is sepcifically targetting that parrot
jonathan My initial thoughts on unifying call_sig and context is "omfg don't do that" 23:57
erm, CallSig
Multi-dispatch needs to make a decision on what to invoke based on the call signature and *then* create a context.
Whiteknight jonathan: why not?
jonathan e.g. the context of the chosen candidate. 23:58
Whiteknight we allocate register storage separately from the Context PMC itself
jonathan At the point we call VTABLE_invoke, we don't know whether we're going to want a new context.
Whiteknight we can create the Context before hand and initialize it when the candidate is found 23:59
jonathan Whiteknight: That feels hacky/icky.
Whiteknight and if we don't use the contexty parts of the new franken PMC, we don't use it
jonathan I'm all for optimizations.
dukeleto darbelo, Whiteknight: those are some better definitions. but I can also see that a single contributor was working on something and then RL got the way for X months
it could still pick back up again
jonathan But this feels like losing a separation of concerns.
dukeleto i think i like the term 'historical' over 'abandoned'

Powered by Cro and the Raku® Programming Language.

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

Thank you!