»ö« Welcome to Perl 6! | perl6.org/ | evalbot usage: 'perl6: say 3;' or rakudo:, niecza:, std:, or /msg camelia perl6: ... | irclog: irc.perl6.org | UTF-8 is our friend!
Set by sorear on 25 June 2013.
dalek c: fbacd28 | Mouq++ | / (2 files):
Add html/syntax for misc definitions
 00:17
dalek c: c473a1b | Mouq++ | / (5 files):
Basic support for definitions via X<>
 01:08
rurban parrot will rename ops2c to parrot-ops2c in the next release. see github.com/parrot/parrot/issues/900 02:05
dalek c: ba47ea5 | Mouq++ | htmlify.p6:
Restructure find-definition to allow for multiple definitions
 02:38
dalek p: 6ab57d7 | rurban++ | tools/ (2 files):
prepare for new parrot-ops2c

with fallback if config{ops2c} does not exist, < parrot-6.9.0		 02:44
dalek c: 60cdbe9 | Mouq++ | / (2 files):
$dr -> $*DR

Simplifies code and doesn't seem to have any preformance penalties		 04:47
dalek p: 2b5b400 | rurban++ | tools/lib/NQP/Configure.pm:
fixup parrot-ops2c for rakudo

use the parrot:: prefix, otherwise rakudo verify_install will fail		 05:13
dalek ast: 26af8f8 | usev6++ | S09-typed-arrays/arrays.t:
Add test for RT #114968
 06:27
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=114968
jnthn morning, #perl6 06:57
lizmat morning jnthn 06:58
jnthn sips coffee and slowly wakes up
bartolin m: my $x = 'Just Another'; my $y := $x; say $y; 07:10
camelia rakudo-moar bc6de5: OUTPUT«Just Another␤»
bartolin say $y; ## is this expected behaviour in REPL?
m: say $y; ## is this expected behaviour in REPL? 07:11
camelia rakudo-moar bc6de5: OUTPUT«===SORRY!=== Error while compiling /tmp/B6ZGPfTKe4␤Variable '$y' is not declared␤at /tmp/B6ZGPfTKe4:1␤------> say $y⏏; ## is this expected behaviour in REPL␤ expecting any of:␤ postfix␤»
moritz bartolin: looks like a bug to me
bartolin m: say $x;
camelia rakudo-moar bc6de5: OUTPUT«===SORRY!=== Error while compiling /tmp/WFxWXUlZt_␤Variable '$x' is not declared␤at /tmp/WFxWXUlZt_:1␤------> say $x⏏;␤ expecting any of:␤ postfix␤»
moritz (in the REPL, doing a 'say $x' on a separate line gives (Mu)) 07:12
bartolin moritz: locally I get 'Just Another' for 'say $x', but (Mu) for 'say $y' 07:13
moritz bartolin: right
bartolin: (camelia doesn't use the REPL, so you can't reproduce it here)
bartolin: that's a bug
bartolin moritz: oh, I see (camelia doesn't use the REPL) 07:14
moritz: okay, I'll submit a rakudobug.
moritz it spawns a new process for each line
bartolin makes sense 07:15
moritz everything else would cause too much confusion in the general case
lizmat jnthn moritz: parrot is currently barfing on $!PIO := nqp::getstdin();
is there something wrong with that ?
jnthn lizmat: Not that I know of 07:16
lizmat does this ring a bell" "Unmarshallable foreign language value passed for parameter 'arg'"
moritz uhm 07:17
lizmat: does it go away if you explicitly type $!PIO as Mu 07:18
?
istr that we had to do something like that in the early days of the nom branch
lizmat tries that
I seem to recall Mu on attributes in setting didn't work 07:19
we'll know in a few mins
hmmm... I replaced nqp::bindattr(self, IO::Handle, '$!PIO', with $!PIO := 07:21
maybe on parrot, that is *not* the same ??
jnthn I'm not sure. I do know that the Parrot code-gen stuf pre-dates the JVM and Moar code-gen by some way 07:22
And while the JVM and Moar ones are quite consistent with each other, there are discrepancies with how things are on Parrot, typically around boxing semantics.
lizmat Mu didn't fix it 07:23
going for nqp::bindattr(self, IO::Handle, '$!PIO', reversal now
jnthn ok 07:24
jnthn ain't sure exactly how you could get that error out of it... 07:25
lizmat it's the only functional difference with the old code that I can see 07:27
jnthn k
lizmat building again, will know result in a few mins
jnthn OK. With luck my last couple of students will be here by then and I'll be attemtping to teach useful concurrency stuff :) 07:28
yup :) 07:29
lizmat no difference :-( :-( 07:31
lizmat seems parrot was barfing over the use of given / when 07:50
FROGGS in combination with a capture?
lizmat given $!path.what {
when << <STDIN> >> {
$!PIO := nqp::getstdin();
FROGGS because I think the error also pops up when you pass a non-sixmodelobject to a capture or so 07:51
ahh, hmmm
lizmat reverting the bindattr change now, see if that makes a diff
FROGGS sadly that one has to wait soo long when testing this :/ 07:52
lizmat yes, meanwhile I'm doing other things
exercise in concurrency :-) 07:53
m: say 1 & say 2 & say 3 # this I expect
camelia rakudo-moar bc6de5: OUTPUT«3␤all(2, True)␤all(1, True)␤»
lizmat m: say 1 S& say 2 S& say 3 # this I find strange
camelia rakudo-moar bc6de5: OUTPUT«3␤all()␤all()␤»
lizmat S03:1107 would indicate sequential evaluation for S& ? 07:55
synopsebot Link: perlcabal.org/syn/S03.html#line_1107
moritz precedence? 07:57
m: (say 1) S& (say 2) S& (say 3)
camelia rakudo-moar bc6de5: OUTPUT«1␤2␤3␤»
lizmat ah, good point :-)
bartolin m: S05 says that $/ is implicitly lexically scoped. does that mean that the following two pieces of code should behave identically? 08:02
camelia rakudo-moar bc6de5: OUTPUT«===SORRY!=== Error while compiling /tmp/H6kqLpK6YS␤Two terms in a row␤at /tmp/H6kqLpK6YS:1␤------> S05 says that $/ ⏏is implicitly lexically scoped. does tha␤ expecting any of:␤ postfix␤ i…»
bartolin S05 says that $/ is implicitly lexically scoped. does that mean that the following two pieces of code should behave identically?
m: { my $x = "foobar"; my $/ := $x; say "INNER: $/"}; say "OUTER: $/"
camelia rakudo-moar bc6de5: OUTPUT«INNER: foobar␤OUTER: ␤»
bartolin m: { my $x = "foobar"; $/ := $x; say "INNER: $/"}; say "OUTER: $/"
camelia rakudo-moar bc6de5: OUTPUT«INNER: foobar␤OUTER: foobar␤»
FROGGS bartolin: in your last example the $/ defaults to the outer (which still is lexically scoped) 08:04
dalek kudo/nom: fcf6dcf | (Elizabeth Mattijsen)++ | src/core/IO/Handle.pm:
Unbreak rakudo build on parrot

It looks like parrot gets thoroughly confused by using smart match in this way.		 08:07
lizmat at least, when used in the setting
bartolin FROGGS: okay, thanks. so in this example $/ within the block is behaving just like an ordinary variable?
FROGGS bartolin: yes
m: sub foo { say $/ }; foo() # it is just implicitely declared 'my' in such situations... 08:08
camelia rakudo-moar bc6de5: OUTPUT«Nil␤»
FROGGS like packages or subs (or units?)...
rurban lizmat: looks like a nqp-p problem to me with ~~
lizmat yeah anyway, it builds again 08:09
spent way too much time on that already :-(
rurban thanks!
lizmat you're welcome: I broke it, so I should fix it 08:10
it's just something I did not expect :-(
bartolin FROGGS: okay. then probably tests like that for RT #72956 (in S05-match/capturing-contexts.t) would like to use 'my $/' within their block, right. 08:11
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...l?id=72956
bartolin m: {$/ := 'foobar';} say $/; 08:13
camelia rakudo-moar bc6de5: OUTPUT«===SORRY!=== Error while compiling /tmp/ib9oq8viXs␤Two terms in a row␤at /tmp/ib9oq8viXs:1␤------> {$/ := 'foobar';} ⏏say $/;␤ expecting any of:␤ postfix␤ statement end␤ statement…»
bartolin m: {$/ := 'foobar';}; say $/;
camelia rakudo-moar bc6de5: OUTPUT«foobar␤»
bartolin m: {my $/ := 'foobar';}; say $/;
camelia rakudo-moar bc6de5: OUTPUT«Nil␤»
dalek kudo/nom: f573ede | (Elizabeth Mattijsen)++ | src/core/IO/Handle.pm:
First part of cruft removal from IO::Handle

Because before, .IO would give you an unopened IO::Handle. So any time you would call a method on an IO::Handle that would require an open file, it would have to check whether the file was open already or not, and whether it was a directory or not. Because it is now a lot harder to create an unopened IO::Handle, we can drop the tests now.
This should particularly make loops with .get and .getc a *lot* faster.		 08:38
lizmat argh, now looking at some tests that segfault on parrot 08:53
dalek kudo/nom: 65819d1 | (Elizabeth Mattijsen)++ | src/core/IO/Handle.pm:
Eradicate the IO::Handle "isDir" attribute
 08:57
carlin are those isDir changes going to cause regressions for RT #122358 ? 09:03
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=122358
dalek ast: 5fe25b6 | usev6++ | S0 (2 files):
Modify and unfudge tests for binding $/, add test for RT #114726

but binding to the outer $/ would lead to problems with later tests hence the modification to use "my $/ :=" instead of "$/ :="		 09:06
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=114726
lizmat it shouldn't, as slurp() uses IO::Path.slurp, not IO::Handle.slurp
arguably, slurp and spurt on IO::Handle should go 09:07
a bare slurp() might, as that uses $*ARGFILES... hmmm... 09:08
bartolin it would be great if someone could check whether my last commit to roast makes sense. (all tests pass, but maybe I missed the purpose of binding to non local $/ in the formerly skipped tests) 09:10
lizmat no, it shouldn't
but I will need to fix ArgFiles.slurp before I can get rid of IO::Handle.slurp :-)
Ven S06:L1628 looks like it has a typo (and just below as well) 09:25
S06:1628 ? synopsebot ?
synopsebot Link: perlcabal.org/syn/S06.html#line_1628
Ven thanks
dalek href="https://perl6.org:">perl6.org: 4914096 | (Kamil Kułaga)++ | includes/menu-nav:
#11 link faqfaq.perl6.org/
 09:36
href="https://perl6.org:">perl6.org: b1ba4e1 | moritz++ | includes/menu-nav:
Merge pull request #12 from teodozjan/master

  #11 link faqfaq.perl6.org/
dalek kudo/nom: 9a6501a | (Elizabeth Mattijsen)++ | src/core/ (3 files):
Deprecate IO::Handle.spurt

In favour of IO::Path.spurt. Spurting on an opened handle is not really anything else but print/write on the handle.		 10:32
ast: afbef95 | usev6++ | S05-ma (2 files):
Remove duplicate test for RT #70007
 10:33
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...l?id=70007
bartolin r: say $] ## LTA error message, RT #119293 11:00
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=119293
camelia rakudo-{parrot,moar} 65819d: OUTPUT«===SORRY!=== Error while compiling /tmp/tmpfile␤Unexpected closing bracket␤at /tmp/tmpfile:1␤------> say $⏏] ## LTA error message, RT #119293␤»
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=119293
colomon Fire::Directory::Tree started failing overnight, I'll bet it was lizmat++'s changes to IO 11:02
leont colomon: you shouldn't put trees on fire! ;-) 11:02
colomon whoops, *File 11:03
lizmat goes to fetch and investigate
Ah, I see: tmpdir *was* returning a Str, but now returns an IO::Path 11:06
Str.path gives an IO (for historical reasons, and should probably be deprecated) 11:07
IO::Path.path gives a string, as that is just the $!path attribute really
hmmmm....
moritz though IO::Path.path was a no-op 11:09
that is, returning the invocant 11:10
but that might have changed too
lizmat yes, it has 11:11
suggestions for improvement welcome :-) 11:12
dalek kudo/nom: b3c26ed | (Elizabeth Mattijsen)++ | src/core/IO/Spec.pm:
Deprecate IO::Spec.tmpdir in favour of $*TMPDIR

IO::Spec.tmpdir also now returns a Str, rather than (the recently changed) IO::Path. This unbreaks testing of File::Directory::Tree.		 11:17
lizmat colomon: ^^^
colomon lizmat++ # launching rerun of spec test 11:19
brrt \o 12:01
lizmat brrt o/ 12:04
FROGGS olá 12:06
brrt \o lizmat 12:08
how p6 been doing since.. last week? 12:09
and how are the preparations for APW going along?
colomon lizmat: huh, now OpenSSL is broken instead. :\ 12:13
lizmat goes check 12:14
installs ok here, but I have additional changes 12:15
dalek kudo/nom: fc75710 | (Elizabeth Mattijsen)++ | src/core/ (2 files):
Deprecate IO::Handle.slurp|copy
 12:16
lizmat pretty sure this won't make a change, though
what is the error with OpenSSL (on which OS/VM ?) 12:17
dalek kudo/nom: 0d20bdd | (Elizabeth Mattijsen)++ | src/core/IO/ (4 files):
Give up on IO::Path.all on parrot

Just too many strange errors, segfaults, repr_unbox_whatever errors that just take too much time. If someone wants to use this newer feature on parrot, they are welcome to provide a fix!		 12:32
colomon lizmat: sorry for spacing out there. 12:33
lizmat: rebuilding rakudo, then I'll give it a try directly and see what happens 12:34
lizmat colomon++
lizmat m: say "foo".IO ~~ :d 12:40
camelia rakudo-moar b3c26e: OUTPUT«False␤»
lizmat m: say "foo".IO ~~ :d:r 12:41
camelia rakudo-moar b3c26e: OUTPUT«===SORRY!=== Error while compiling /tmp/y0saYFSJjz␤You can't adverb that␤at /tmp/y0saYFSJjz:1␤------> say "foo".IO ~~ :d:r⏏<EOL>␤ expecting any of:␤ pair value␤»
lizmat would be nice if the above would be syntactic sugar for "foo".IO.all(<d r>)
colomon hmmm, works if I invoke Panda directly.
dalek ecs: 6b22f92 | (Elizabeth Mattijsen)++ | S03-operators.pod:
Don't give people the wrong idea :-)

File tests are primarily for paths, *not* for handles		 12:43
colomon lizmat: either it was just something transient or your change fixed it, because it's working fine for me in the smoker now. 12:44
lizmat ok, good to hear!
arnsholt jnthn: github.com/jnthn/zavolaj/issues/47...t-58010593 looks like a Moar internals thing, no? 12:45
FROGGS arnsholt: likely 12:46
lichtkind is there a shortcut in p6 to rotate values like $a += 2; $a = 0 if $a > 10 ? 12:50
lizmat something with mod comes to mind?
lichtkind ah thats an idea thanks 12:51
lichtkind would even work in p5 12:51
psch hi #perl6 o/
dalek kudo/nom: 0730d44 | (Elizabeth Mattijsen)++ | src/core/IO/Path.pm:
Allow all possible file tests on .chdir/chdir()

Except on parrot, because it doesn't like IO::Path.all, so the old restriction to <r>, <r w> and <r w x> remains there.
arnsholt lichtkind: Maybe %= could work? 12:52
moritz $a = ($a + 2) % 10; 12:55
colomon what moritz++ said. Clean, simple, and straightforward 12:56
lichtkind true :)
lizmat except lichtkind's code said $a > 10 12:57
I guess that would work, yes, as the check is after the addition :-)
lichtkind yeah but i can adapt :) 12:59
colomon sure, so it's % 11
lichtkind it was an example and not with actual numbers i need anyway :) 13:00
lizmat now for an excercise, turn this into a rot10 operator :-) 13:00
using macros :-)
moritz a meta operator, please 13:01
lizmat where the "10" would become one of the parameters to the function
moritz $a = 42 rot10+ 5
$a rot10+= 2;
eeeks
lizmat of course, rotN would do something differently on strings, e.g. rot13 :-) 13:03
lichtkind haha
critcs are right perl ahs builtin encryption :)
colomon wouldn't it be much more sane to define a modular group? 13:04
lichtkind we can stack that metaop on the parser grammer anyway
sure
lizmat
.oO( I still have a Perl 5 IO layer module on CPAN doing rot13, would anybody be interested in maintaining that ? )
lichtkind haha
shit no original idea from me today 13:05
not now
abraxxa arnsholt: hi! 13:09
arnsholt: thanks for your help so far!
arnsholt Not a problem 13:10
arnsholt From looking at the test suite, it looks like it dies in DBIish.install_driver 13:13
Try making a test script with install_driver inlined and see if that explodes with the same error 13:14
colomon gist.github.com/colomon/fbdee769ed36d9836cbb # kind of awkward but works for addition 13:15
arnsholt And perhaps try to eliminate the use of EVAL and just put in the code that's called for the Oracle code
abraxxa: If it still errors out without the EVAL, see if it's the act of using DBDish::Oracle that causes the failure, or the call to DBDish::Oracle.new 13:16
abraxxa arnsholt: ok 13:17
arnsholt Actually, try the use DBDish::Oracle first
The error from serialisation might point in that direction 13:18
abraxxa arnsholt: my $drh = DBIish.install_driver($*mdriver); directly in t/45... throws Serialization Error: Unimplemented case of read_ref 13:21
maybe my DBDish class doesn't implement enough methods to be compatible so far?
arnsholt That sounds weird 13:22
Try to have a file with just "use DBDish::Oracle; say 'alive';" in it
(And the library paths set up correctly, obviously)
abraxxa perl6 -e 'use DBDish::Oracle; say "alive";' 13:23
same
arnsholt Yeah
abraxxa ah, -Ilib works
arnsholt Yeah, that might work
abraxxa now t/45... also executes ok
arnsholt Since that won't use the precompiled modules, but compile it directly on the fly 13:24
abraxxa original t/45 now Segmentation fault (core dumped)
arnsholt So your module somehow triggers an edge case in serialisation/deserialisation
abraxxa so that was the difference between panda install . and using perl6 directly
i thought it's passing a pointer to an unreserved memory segment 13:25
arnsholt Nah
abraxxa the OCIEnvNlsCreate call is fine, if i put a die in the if i get -1 as return value 13:26
arnsholt Yeah, that sounds reasonable
dalek q: e1ed104 | pmichaud++ | answers.md:
s/backtrace/backtrack/ when talking about 'rule' and regexes.
arnsholt If there's an error, the handle probably won't be valid =)
abraxxa arnsholt: but that's how you shoud get the error text according to the oracle oci docs 13:28
arnsholt: docs.oracle.com/cd/E11882_01/appdev...LNOCI16220
arnsholt abraxxa: From docs.oracle.com/cd/E11882_01/appdev...tm#i540516 it looks like you need to use the "environment handle", whatever that is 13:32
abraxxa OCIEnvNlsCreate should return this handle 13:33
arnsholt Oh, that's the thing created by the Create call
Right 13:34
abraxxa i tried OCI_DEFAULT instead of OCI_THREADED, also segfaults
colomon gist.github.com/colomon/3a2861c1b9e492836b75 # hard to recognize Unicode twist
abraxxa do i need to pass the memory allocation functions to it? 13:35
does NativeCall expose them?
arnsholt Oh, I think I see it
Or maybe not
What type does OraText correspond to?
dalek q: e6aa625 | pmichaud++ | answers.md:
A bit more rigor in describing "Any".
 13:36
jnthn arnsholt: Does it work if the module ain't pre-comp'd? It looks like something was written bogus serialization data, or is trying to read it in a bogus way...
arnsholt jnthn: Yeah, sounds like that from what abraxxa says 13:36
abraxxa: It looks like OraText is a character type, from the descriptions in the manpage 13:37
jnthn arnsholt: That error sounds like a silly bug ratehr than a deep one
jnthn arnsholt: As in, something is writing one format and the reader is reading it differently 13:38
arnsholt Right, right
abraxxa arnsholt: yes, that's why I've defined the pointer as my @errortextp := CArray[Str].new; 13:39
arnsholt abraxxa: There's a type mismatch in your use of ErrorGet. The errorcodep argument is declared as OpaquePointer, but your passing a CArray 13:39
arnsholt abraxxa: Yeah, but that array is empty. So it only has a small amount of storage allocated intially (from memory) 13:39
arnsholt And you're claiming that it's 512 bytes, which it certainly isn't 13:39
abraxxa arnsholt: correct, that's why I asked if i have to preallocate it by e.g. filling it with zeros 13:40
arnsholt But I suspect OraText* corresponds to Str, not CArray[Str]
abraxxa arnsholt: but i have to pass a pointer to a Str
arnsholt Well, that depends on what OraText is. Is it a char-like type, or a string-like type? 13:41
abraxxa I've interpreted your response on github that I should use CArray to get a pointer to it
arnsholt Well for a char** type argument that's correct
abraxxa how do I find out the details of OraText?
arnsholt But the description of OCIErrorGet only really makes sense if the argument is char*
abraxxa oratypes.h has typedef OraText *string; 13:42
arnsholt From docs.oracle.com/cd/E11882_01/appdev...#BGBCCBGF: static OraText *username = (OraText *) "HR"; 13:42
It's definitely a character type 13:43
abraxxa paste.scsys.co.uk/428318 13:43
arnsholt Which gets into a different weak point of NativeCall
arnsholt The correct type for that argument is CArray[int8] 13:44
abraxxa i tried int32
nah, that was errorcodep
arnsholt So you'll have to make a CArray[int8] and write a 0 to the element with index 511 (that'll resize and zero all the preceding slots) and pass that in
errorcodep should be CArray[int32] 13:45
abraxxa still segfaults 13:46
Native call expected object with CPointer representation, but got something else 13:47
but it doesn't print which call and arg
arnsholt I think it's errorcodep 13:49
dalek kudo/nom: ef945a7 | (Elizabeth Mattijsen)++ | src/core/IO/Handle.pm:
Some more parameter sanity and cleanup
arnsholt You've declared it as OpaquePointer but you're passing a CArray
abraxxa is my definition wrong or the call?
because I need to access the value the pointer points at 13:50
arnsholt Assuming you actually want to look at the value, it should be CArray[int32]
Which means that both the definition and the call is wrong, since you're actually passing a CArray[Str] =)
abraxxa i've changed both, looks good now 13:51
how do i get at the CArray[int8] errortext now? 13:52
arnsholt Then you copy the contents of the CArray into a Buf, which you can decode (or is it encode, I never remember which is which) to a string 13:53
abraxxa arnsholt: i've tried that in the past, there is a problem with that... 13:54
dalek kudo/nom: 8f625c9 | (Elizabeth Mattijsen)++ | / (4 files):
IO::FileTestable is no more
 13:56
arnsholt You're comparing the errorcode as if it's a string too, BTW 13:57
abraxxa is there a channel log? 13:59
arnsholt irc.perl6.org
moritz yes, at irc.perl6.org
abraxxa we've discussed the Buf problem on the 12th September says my log
starts here irclog.perlgeek.de/perl6/2014-09-12#i_9343864 14:00
Malformed termination of UTF-8 string 14:01
;)
we got an error message
abraxxa regarding OCI_DEFAULT vs OCI_THREADED, which one should I use? I thought for Perl 6 OCI_THREADED would make sense but I'm not sure if that is supported by NativeCall 14:02
arnsholt I have no idea what those mean 14:03
What's the difference between the two modes of operation?
nine Regarding the suggestion of an "a bit more stable than nom" branch. How about calling it "master"? nom would be an integration branch and master a continuously releasable branch like in many other projects. 14:06
arnsholt abraxxa: Anyways, it should work if you copy the int8 elements of the CArray into a Buf[int8] and then decode that into a Str 14:07
abraxxa arnsholt: I 'guess' threading support
moritz nine: release/master or master/dev would be a good pair
arnsholt From the usage example of OraText I linked earlier, it's clearly not a UTF-16 string
arnsholt abraxxa: In that case, I'd go with default until you know what threaded mode does =) 14:08
abraxxa ok
[Coke] japhb: so, the issue with the daily runs keeping track of what's the latest stable build: roast. We don't have a way to say "this passes all spectests at <rev>", just "at some point". 14:08
abraxxa arnsholt: see newest commit 14:09
let me look up why i encode this as UTF-16
nine abraxxa: I'm curious. How's your Oracle driver coming along? 14:10
abraxxa nine: GAH!
:P
I found a poor soul with github issue (arnsholt) who helps me understand
arnsholt++
arnsholt That looks more reasonable 14:11
abraxxa i guess it was a misinterpretation of OraText
arnsholt Yeah, OraText is a terrible (TERRIBLE!) name for it
So you get a sane error message out now? 14:12
abraxxa can i define OraText in the perl 6 module with the same name as CArray[int8] to use the same name everywhere?
'Malformed termination of UTF-8 string' is the error text
so how do i have to pass the username and password to OCILogon2? 14:13
no, that can't be the problem as that happens later
arnsholt The malformed termination error comes from the call to NlsCreate? 14:14
abraxxa AH
docs.oracle.com/cd/E11882_01/appdev...tm#i562098
read charset(IN)
i'm passing OCI_UTF16ID 14:15
abraxxa that's why I'm encoding the username and password to UTF-16 later down 14:15
arnsholt Right. Just make it UTF-8 instead 14:16
That'll probably save you some grief in the end
abraxxa arnsholt: not supported 14:18
arnsholt: i'm not so sure any more @error 14:19
i wonder if i have to use OCINlsCharSetConvert everywhere 14:21
arnsholt Could be 14:22
I've absolutely no idea how the OCI API works 14:23
abraxxa arnsholt: is that the correct invocation? github.com/abraxxa/DBIish/blob/mas...e.pm6#L467
arnsholt: me neither
arnsholt It looks reasonable, but I'm not entirely sure how the Buf API works
abraxxa can the error come from this call? 14:24
arnsholt Entirely possible. Try putting a say on either side of it 14:25
dalek ast: 938728b | (Elizabeth Mattijsen)++ | S32-io/dir.t:
Esthetics
ast: 2c843cc | (Elizabeth Mattijsen)++ | S32-io/spurt.t:
Properly plan and fudge for niecza
arnsholt If only the one is called, it's Buf that's unhappy
It could be that it doesn't like getting the CArray as an argument
abraxxa yeah, coming from decode 14:26
arnsholt Try copying over the elements one by one
arnsholt r: for 0..^3 { .say } 14:26
abraxxa arnsholt: that should be fine according to doc.perl6.org/routine/new#class_Buf
camelia rakudo-{parrot,moar} 0730d4: OUTPUT«(timeout)» 14:27
arnsholt abraxxa: Try copying over the bytes one by one, but stopping when you reach a NULL byte 14:28
It might not like the trailing zeroes 14:29
abraxxa arnsholt: is there a helper function available in NativeCall? 14:30
arnsholt Oh. It might be that the resize doesn't actually zero the array 14:31
I thought it did, but now that I think about it, I don't think it does
abraxxa DBDish::Pg uses CArray[Str], maybe there is support for this already?
github.com/abraxxa/DBIish/blob/mas...Pg.pm6#L22
arnsholt Try setting all the elements to 0
No, there's no support for this particular use case 14:32
It's something we've discussed in the past, but we haven't really figured out how to best handle it yet
Or actually
You could also get away with copying from the CArray one by one and stopping at the first NULL byte
arnsholt Which you'll probably end up doing anyways, since if it's nulled, you'll get a Str with a lot of trailing nulls 14:33
laben hai #perl6 o/ 14:34
abraxxa arnsholt: what's the neatest syntax of setting all 512 int's to 0? 14:35
arnsholt Neatest syntax is 0..^512, IIRC 14:37
sjn \o
arnsholt Fastest syntax is probably while $i < 512
for 0..^512, that is
xfix And 0..^512 can be shortened to ^512, right? 14:38
masak hejhej, #perl6
abraxxa but i'd need the iterator number for the array access, or should i use push?
masak xfix: right.
moritz ho ho masak
lizmat masak moritz o/
masak moritz: ho ho ho!
arnsholt xfix: Yeah, I think so 14:39
abraxxa: Either way probably works
dalek p-js: 7f00d5e | (Pawel Murias)++ | src/vm/js/QAST/Compiler.nqp:
Remove misleading comment.
 14:41
pmurias masak: hi
laben lizmat: i have something strange going on, can you help me?
lizmat sure
laben m: my @a=1..5; for @a { indir "foo", { say "OK"; }} 14:42
camelia rakudo-moar 0730d4: OUTPUT«(timeout)»
laben this errors out
oops i put m:
abraxxa Cannot call 'push'; none of these signatures match
@errortextp.push(0) for ^512;
laben it says Unknown tesst r w
abraxxa it seems CArray doesn't support push 14:43
lizmat laben: what VM ?
laben moarvm ofc
moritz @errortextp[$_] = 0 for ^512;
laben i pulled 2h ago
arnsholt Oh, for CArray you're stuck with array indices, yeah
lizmat laben: confirmed
arnsholt push isn't there because for a pointer returned from C we don't know anything about how many elements there are or how much room is available 14:44
abraxxa moritz: thanks 14:45
masak Mouq: if you example irclog.perlgeek.de/perl6/2014-10-05#i_9460255 -- was it deliberate to do `sub where` rather than `macro where`? you seem to return a quasi at the end, just like from a macro. 14:45
lizmat laben: fixing now 14:46
laben lizmat: perfect :-) it's great having such active devs
abraxxa arnsholt: the values aren't overwritten by the call but still both contain zeros 14:47
laben btw, "filename".IO.slurp and slurp "filename" both return deprecation warnings now, what is recommended? 14:48
arnsholt abraxxa: Not sure I understand what you mean
abraxxa after calling OCIErrorGet the values still only contain the zeros i"ve prefilled them with
arnsholt Huh. That's odd 14:49
Not sure what's going on in that case
lizmat "filename".IO.slurp shouldn't give a deprecation warning, not should slurp "filename|" ? 14:50
hmmm...
dalek kudo/nom: 9818ba1 | (Elizabeth Mattijsen)++ | src/core/IO/Path.pm:
Fix issue with indir, spotted by laben++
 14:52
laben lizmat: i'll rebuild and retest ASAP 14:53
xfix I wonder how easy would it be to create a domain specific language in Perl 6 that would allow you to write code like `if (SELECT first_name || ' ' || second_name AS $name FROM users WHERE id = 1) { say "User #001's name is $name." }`. 14:55
moritz shouts "macros", just to annoy masak++ :-) 14:57
laben lizmat: indir fixed, many thanks to lizmat++ 15:04
deprec warnings remain though
lizmat it was only a 1 character change :-)
yes, working on that
masak moritz: probably more of a slang, but time will tell ;) 15:10
moritz masak: I think so too
masak: but it's more fun to annoy you with "macro!" :-) 15:11
jnthn moritz: no no, to annoy masak, should MARCO!
masak auugh
moritz jnthn: macro polo!
jnthn swiftly decommutes :)
pmurias somebody should write a marco module that allows users to write 'use marco; marco foo {...}' ;) 15:16
masak "defining keyword synonyms". got it. 15:19
lizmat Mu $!path Mu $!PIO int $!ins Mu $!chomp 15:21
P6opaque: no such attribute '$!PIO'
hmmm.. something wrong in this picture 15:22
japhb [Coke]: Part of the reason I was thinking of the daily runs is that we could time them to run during the lull when Europe is mostly asleep, when commits are mostly idle. And that might put gentle pressure on people not to push a major refactor to nom just before going to sleep, analogous to at most companies not pushing new software releases on Friday evening. 15:25
timotimo so much backlog i just skipped because i don't want to touch anything related to the oracle database ... :P
carlin just think about how much Larry Ellison's networth increased by while you were scrolling past it 15:28
abraxxa x 15:32
dalek kudo/nom: c24537e | (Elizabeth Mattijsen)++ | src/core/IO/Path.pm:
Fix deprecation warning on IO::Path.slurp
lizmat laben: ^^^
lizmat cycling& 15:32
abraxxa arnsholt: the problem was that I defined $envhp as OpaquePointer and assigned @envhpp[0] to it hoping that I later can use $envhp which isn't the case 15:33
laben lizmat: retest coming right up
abraxxa instead I have to assign @envhpp[0] to $envhp to get the filled pointer out 15:34
abraxxa arnsholt: ORA-24483: Invalid parameters were passed into environment creation call 15:41
YEAH
abraxxa how do i pass null to NativeCall? 15:45
timotimo just use OpaquePointer 15:51
timotimo any type object will be treated as NULL, but OpaquePointer's type object seems like a common one to use 15:52
laben lizmat: sorry i did not answer quickly, i got a little busy. Everything is fine now, no deprec warnings :) 16:02
abraxxa timotimo: thanks 16:10
passing OpaquePointer instead of a CArray[OpaquePointer] fixed the OCIEnvNlsCreate call 16:13
[Coke] japhb: I think having something like this is a good idea; I'll play with it and let you know. Maybe good enough will be good enough. 16:18
japhb [Coke]: OK, cool, thank you. 16:21
tony-o carlin: np, thank you for creating the merge req 16:23
psch does --no-regex-lib in the invocation for compiling NQPHLL.nqp mean that i don't get any regex and thus no grammars? 16:41
...and would that segfault if i use a grammar anyway?
pmurias you don't get any regex or grammar support
pmurias it's not supposed to segfault (haven't actually tried it so it might) 16:42
colomon ponders planned segfaults...
psch pmurias: thanks. i suppose the segfault isn't really much of a problem if i'm doing something that shouldn't be done.
i guess i'd have to get parsing of command line options out of NQPHLL-land then 16:43
i also wish i'd have asked earlier, heh
but then moritz++ did raise a good point with 'multiple -e might not really make sense and people should write a proper script when it gets complicated' 16:44
i'll repurpose my nqp PR for multiple -e towards "forbid multiples of options we definitely don't want more than one of" i think
PerlJam multiple -e might make sense in conjunction with multiple -I 16:45
pmurias multiple -I is something we really want 16:46
PerlJam for a completely made-up example: perl6 -e 'perl 6 here' -IInline::Python -e 'python here' -ISome::Other::Slang -e 'you get the idea' 16:47
er, ... sorry that should have been -M, not -I 16:48
my brain or something isn't quite working right.
PerlJam (But, yeah, multiple -I is certainly wanted :) 16:50
masak PerlJam: how does Perl 5's `perl` work in that regard? it already has both multiple -M and multiple -e 17:09
tony-o i use multiple I with p5 pretty regularly, i just did it to time min/max vs single loop
leont perl 5 has use semantics for -M 17:12
psch masak: gist.github.com/peschwa/0864c6a62a16218a17a8 17:14
-Ms seem to get prepended
at least if O=Deparse isn't cheating more than it's supposed to
tony-o interesting
masak psch: very interesting. 17:19
given what I know of the cmdflags codegen, this makes sense. 17:20
moritz rakudo currently stops parsing after the first -e 17:31
which IMHO makes a lot of sense, because everything after going into @*ARGS is a very simple and natural rule 17:32
multiple -e would imply a more complicated rule for deciding when to stop 17:33
and until I've seen a good use case for multiple -e, I'm against more complicated rules
I've nobody objects, I'll patch S19 accordingly
PerlJam so, that implies that -e must be the last option? 17:34
moritz yes
PerlJam or, that -e is effectively a stopper for future options. (like -- normally does)
wfm I guess. I have no real need of multiple -e
moritz those two are equivalent, no?
tony-o i treat them that way but i think it'd be confusing to people who don't 17:35
tony-o like the argument ordering in OSX vs linux 17:35
moritz who are arguments ordered in OSX? 17:36
tony-o rm folder -Rf in OSX won't remove a folder 17:36
it won't parse -Rf until after it tries to remove the folder 17:37
abraxxa arnsholt: it starts to work! look at my latest commits if you like
tony-o same argument order works fine in gnu 'rm' which is what most linux distros come with iirc
moritz tony-o: so 'rm a -Rf b' would remove b/ but not a/ ?
tony-o: or would it try to rm files 'a', '-Rf' and 'b' ?
tony-o it errors out as: rm a: is a dir rm: -Rf: no such file or directory 17:38
moritz so, the latter
ok, thanks
tony-o same error with b, it treats the non flag argument as the end of flags
masak PerlJam: or alternatively, that -e '...' has the same effect on (wrt later arguments) as <scriptfile> does
moritz: +1 to only one -e 17:39
s/on /
tony-o +1
PerlJam moritz++ 17:40
psch i don't have a particular use case that makes -e very useful. if anything the one PerlJam brought up before is the only one. 17:42
additionally, i feel like my solution would need much more time to get working properly, which i'll probably be lacking in the near future 17:43
dalek kudo-star-daily: a210212 | coke++ | log/MoarVM-version.log:
today (automated commit)
 17:46
kudo-star-daily: 7bdc24f | coke++ | log/MoarVM-version.log:
today (automated commit)
rl6-roast-data: 9180fe1 | coke++ | / (4 files):
today (automated commit)
rl6-roast-data: 70cccd9 | coke++ | / (4 files):
today (automated commit)
[Coke] I never thought -e was a stopper for args. 17:47
moritz [Coke]: what did you think was a stopper for args?
[Coke]: also, current rakudo already implements -e as a stopper. Did that ever surprise you? 17:48
[Coke] moritz: -- 17:50
moritz: hasn't come up 17:51
moritz [Coke]: so in the case of perl -e 'program' more stuff here, the "more stuff here" is always part of the command line arguments to perl, not the program? 17:51
[Coke] also, it doesn't seem true: "perl -E 'say 3' -d" puts me in the debugger. 17:52
(same with -e) 17:53
moritz [Coke]: rakudo currently implements -e as a stopper, not perl5
[Coke] I thought the argument was "we're doing this because perl 5 does." 17:54
my bad.
moritz we're doing that because
dalek c: 4b84000 | Mouq++ | lib/Language/functions.pod:
Stub functions.pod
moritz 1) p5's disambiguation means that some arguments are interpreted as options to p5, and some as options to the program 17:55
2) those rules change when there's a new command line option
and
3) they are much more complicated than they need to be
[Coke] I will make my -1 tentative instead of firm, then. :) 17:56
moritz noted, thanks
Mouq masak: Yes, I meant macro. And I was fighting the syntax in that example, but I was trying to express some sort of macro for pattern-matching 17:57
nine If perl6 should one day replace perl even for Perl 5 programs, it should be as compatible as possible. 17:59
PerlJam Doing the conservative thing first, then opening it up in the future is far easier than if it were the other way around (ie. we decided after allowing multiple -e that we really only want one and for it to be a stoppper for args) 18:01
nine: I think when that day comes, compatibility with Perl 5 probably won't matter :)
nine PerlJam: you seem to think that this day is far in the future? 18:02
masak Mouq: I am intrigued by your example. 18:03
PerlJam nine: I dunno, but judging by the time scale of previous Perl cultural shifts, yes. 18:04
masak Mouq: could you maybe give an example input, too? i.e. how would that macro be used? the more concrete, the better.
nine PerlJam: I don't see a cultural shift needed. I'd say lower start up time, a stricter adherence to S01 and automatic loading of Inline::Perl5 is all it needs. And those are not that far out. 18:05
Mouq masak: It came up in another project where I wanted to do something similar, although I think in both cases it runs into the issue with blocks in quasis 18:06
masak yes, that's what I thought
Mouq masak: I have class soon, but I'll try to write something up soonish 18:06
masak \o/
Mouq masak: :)
\o
masak quite often I find myself wishing that a macro could flaunt TTIAR a little. 18:06
tadzik heheh
don't we all :)
PerlJam nine: The cultural shift is convincing p5p that that's all it needs (or that the role of p5p changes)
masak I guess that comes down to macros being more grammar-y than they are now. 18:07
dalek ast: cb72de7 | usev6++ | S32-exceptions/misc.t:
Add test for RT #71800
 18:39
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...l?id=71800
dalek nda/CPAN: 29e0bcb | (Elizabeth Mattijsen)++ | lib/Panda/Fetcher.pm:
Fix Panda breakage caused by S16|S32/IO refactor
 19:01
panda/CPAN: 7398d28 | (Elizabeth Mattijsen)++ | / (3 files):
panda/CPAN: Use $*CWD instead of cwd()
FROGGS (not to worry, that's just a branch merge) 19:01
dalek ecs: 8f8c840 | moritz++ | S19-commandline.pod:
[S19] simplify rules for command line parsing

  -e now stops command line parsing. This means only one -e is now interpreted.
Use proper scripts instead.
Rationale: p5's disambiguation rule is that everything after the -e program is considered an option to perl if it's a valid perl option. This means that 1) you have to memorize an arbitrary list of options that apply to the compiler to know where an option belongs to 2) introducing an option to the compiler is a backwards incompatible change 3) the rules are more complicated		 19:06
lichtkind good night 20:15
japhb moritz: FWIW (after you made the spec change): 1. I've actually used the multiple -e with perl5 before. 2. IIRC this was to get around some confusing bit of shell parsing. 3. The benefit gained over just breaking down and writing my code to a file was not, on balance, worth the ugliness in the default option parsing. So in other words, as someone who has used the feature you are declaring dead, I think you made the right decision. 20:26
moritz japhb: thanks 20:29
leont is just at horror at the general behavior, and wonders if it can break MakeMaker (which is using -e + @ARGV in many of the ExtUtils::Command lines) 20:42
leont I'm looking for a roommate in Salzburg, anyone up for that? 20:45
dalek ast: 059caa0 | (David Warring [email@hidden.address] | / (2 files):
snick in a test for RT #122896
 21:01
synopsebot Link: rt.perl.org/rt3//Public/Bug/Displa...?id=122896
timotimo nine: is it safe to say that there's not that much to report on Inline::Perl5 because it's already so close to being done? and you're now spending time on the secret thing you want to unveil at APW? 21:04
japhb nine: And as a followup to timotimo's question, what is *not* done for Inline::Perl5? 21:27
lichtkind there was some talking about closing specs in 2014 21:55
is it still relevant? 21:56
or is it next chsitmas :)
christman
lizmat It is my goal to get clarity about that at the APW 21:57
lichtkind allright so i delete this sentence 21:58
thanks 21:59
sorry cant make it to apw
japhb As a mantle developer, it does not appear to me that Rakudo's core functionality is sufficiently complete to declare specs closed.
lichtkind still 2 exams left
lizmat japhb: clarity does not necessarily equate to closing the specs
for me, it will equate to specifying things we want *for sure* in 6.0 and marking them as such in the spec 22:00
japhb lizmat: I don't disagree there. I'm saying more along the lines of "I mostly expect the clarity to be 'We can't close the specs yet.'"
lizmat I would take such an outcome as not meeting my expectation 22:01
japhb I cautiously accept your optimism. ;-) 22:03
timotimo tony-o: just looking at your DB::ORM::Quicky; you're doing $username ~= chr((65 .. 90).pick) for 0 .. 10; which i think is more simply written as my $username = [~] ("a".."z").roll(10)
lizmat japhb: to me it's more important that we set what we want for 6.0, and that we can tell the world what that would be 22:06
tony-o timotimo: nice, thank you! 22:07
dalek kudo/nom: 26e285b | (Elizabeth Mattijsen)++ | src/core/IO/Spec.pm:
Deprecate what we can from IO::Spec
tony-o m: my $username = [~] ("a".."z").roll(10); say $username; $username = [~] ('a'..'z').roll(10); say $username; 22:09
camelia rakudo-moar c24537: OUTPUT«kkxtznlqji␤wnsmvkhrjo␤»
japhb lizmat: I just haven't seen evidence that any of our previous predictions about solid parts of the spec has actually turned out to be that way *once we actually implemented those parts*
(And I use 'our' loosely here, because I've not made such a prediction that I can recall -- for reasons that recurse with my above statement.) 22:10
lizmat well, things have been moving very fast the past 6 months 22:11
japhb is of the "Versioned specs are what happens after unversioned specs get implemented" school of thought, rather than the "Versioned specs are what comes out of committee and gets implemented" school 22:12
lizmat: Oh sure, and that's *fantastic*.
tony-o timotimo: as soon as i figure out why i have the wrong version of m-CORE i'll push the change
japhb (everyone who committed this year)++ for that
lizmat :-) 22:13
japhb finds this reminiscent of the w3c specs that got enshrined in versioned spec and never came to be -- much less came to be in the same form
timotimo tony-o: you will need to reinstall/recompile all modules after you've rebuilt your rakudo 22:14
lizmat let's not make the same mistakes :-)
tony-o is there a quick way to do that with panda?
lizmat runs a rebootstrap.pl to make sure she didn't break panda
japhb lizmat: *chuckle*. Yes, well, ... I'm aiming for mistake #0. :-)
tony-o really tired of vendors who roll their own "secure" VPN software 22:15
s/vendors/customers/ 22:16
timotimo "rebootstrap.pl" will remember what packages you've had installed and will install them again 22:19
japhb tony-o: <sarcasm>But it's just so *easy* to do, why wouldn't everyone make their own? After all, they can add all the extra security ideas they had back in college, plus use one-time pads for everything! What could go wrong?</sarcasm> 22:21
.oO( Reviewing C++ commits is making me snarky ... )
timotimo japhb: "and use a one-time pad for everything!" :P 22:22
tony-o i normally don't mind because i can do p6 stuff while the customer fixes their half working VPN stuff but this one wants to do all of this mobile stuff with phones and ipads so they can get their data without being on the network but their VPN doesn't work with mobile devices (inc windows phone), or OSX, or linux, or even firefox|chrome|opera|safari|ie10+ on windows 22:24
timotimo m( 22:25
tony-o guess it's better than no customer 22:27
timotimo except if you're expected to fix such things for them 22:28
tony-o hah 22:31
tony-o timotimo: pushed changes 22:39
timotimo posted the weekly, too 22:54
tony-o link?
timotimo p6weekly.wordpress.com 22:56
tony-o oh cool, i made one :-) 23:05
is p6 represented at yapc? 23:07
japhb Which YAPC?
tony-o whichever happens in the US 23:08
i think it was in texas this year
japhb That was last year (I went to that one). It was in Orlando this year, IIRC. 23:09
tony-o yea i'm seeing that now -
japhb I dunno where the next one is.
tony-o i think i'm gonna try to go this year even though perl has nothing to do with what i do for a living 23:10
japhb It's usually announced at the conference.
tony-o this coming year*
japhb nodnod
woolfy1 The next YAPC::NA will be in Salt Lake City in Utah. 23:18
On 8-10 June, 2015. 23:19
Perl 6 was represented at this year's YAPC::NA, and last year's YAPC::NA, and the year before that too... 23:20
timotimo is the next yapc::eu location known yet?
woolfy1 Yep, YAPC::EU in 2015 will be in Granada, Spain, in september 2015.
Exact date... not know yet. 23:21
But september. Not august.
If you want a lot of Perl 6, you gotta go this week to Salzburg, Austria, for the Austrian Perl Workshop & Hackathon. act.useperl.at/apw2014/ 23:24
Lots of Perl 6 people at the hackathon. It is going to be great!
tony-o austria would be a lot of fun 23:25
woolfy1 Salzburg is one of the most beautiful cities on this planet! 23:27
japhb Wish I could be there.
tony-o füssen in germany has been my favorite so far 23:28
woolfy1 Füssen, Bavaria, Germany: population of 14,631; town is known for its violinmaking industry. :-) 23:29
tony-o i stayed there for a week when i was 18 or 19, i like the solitude and german folks 23:30
leont What does the jit-by-default mean for rakudobrew? Is the moar versus moar_jit distinction still relevant? 23:31
timotimo i think we'll want to turn that into "moar" and "moar_nojit" 23:32
as it stands, the "moar" target will give you a moar-jit and "moar_jit" will give you a "--enable-jit is no longer needed" message

Powered by Cro and the Raku® Programming Language.

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

Thank you!