Perl 6 language and compiler development | Logs at colabti.org/irclogger/irclogger_log/perl6-dev | For toolchain/installation stuff see #perl6-toolchain | For MoarVM see #moarvm
Set by Zoffix on 27 July 2018.
Geth nqp: 12998693c9 | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/build/MOAR_REVISION
[MoarVM Bump] Brings 3 commits

MoarVM bump brought: github.com/MoarVM/MoarVM/compare/2...g727767006 727767006 Make release.sh more clear as to which we want written to MANIFEST 86ba33728 [Spesh] Change MVM_VECTOR to alloca da9123a76 Only mark stolen deopt ann used if it is
00:08
¦ nqp: version bump brought these changes: github.com/MoarVM/MoarVM/compare/2...g727767006
rakudo: 4f027ff631 | (Aleks-Daniel Jakimenko-Aleksejev)++ | tools/build/NQP_REVISION
[NQP Bump] Brings 13 commits

NQP bump brought: github.com/perl6/nqp/compare/2018....g12998693c 12998693c [MoarVM Bump] Brings 3 commits 47271b2d2 [jvm] Implement nqp::istype_nd 6ee956f41 [js] Compile qrusage correctly, it still needs to be fixed to use a working dependency ca7e082fc [js] Fix nqp::getsignals to make it box stuff in the array it returns ... (9 more lines)
¦ rakudo: version bump brought these changes: github.com/perl6/nqp/compare/2018....g12998693c 00:09
MasterDuke it's now a bit more generic and actually does some error checking and passes a spectest. all that's left is getting the different math types implemented 03:13
but first comes sleep... 03:14
AlexDaniel .tell Zoffix this list is not perfect and probably needs to be tweaked, but it's better than nothing (single apt command on the bottom of the page): github.com/perl6-community-modules...aster/wiki 03:27
yoleaux AlexDaniel: I'll pass your message to Zoffix.
AlexDaniel .tell Zoffix I think there are some packages from `testing` and `unstable`, and even with these repos added I see libcurl refusing to install… just skip packages that do not want to install 03:42
yoleaux AlexDaniel: I'll pass your message to Zoffix.
AlexDaniel I'm running toaster right now with some deps installed 03:45
who knows, maybe I'll find something :)
AlexDaniel .tell Zoffix “184 out of 1127 (16.33%) appear to be unusable” that's an improvement :) 06:28
yoleaux AlexDaniel: I'll pass your message to Zoffix.
AlexDaniel “If you do not want to spend any more time on this module, could you consider removing it from the ecosystem? Thank you for trying Perl 6!” 06:30
:D
stmuk dtrace on OS X is virtually unusable since it doesn't allow tracing of shell scripts for "security reasons". I saw claims it was possible to turn this off via a complex sequence of steps but never got it to work 09:36
stmuk I think if your shebang line pointed at your own path to a non-system bash it would work 09:36
stmuk but life is too short. This is one of the reasons I don't use OS X anymore 09:37
AlexDaniel eh toast.6lang.org/ 10:40
I'm running it again because there are a lot of flops for some reason
AlexDaniel but I see some modules actually failing now with “Cannot assign to a readonly variable or a value” 10:42
anyway, will check them all after I wake up
AlexDaniel also there are false Succs with double free or corruption on 2018.06, these are really weird 10:46
stmuk AlexDaniel: which version of moar do you toast against? the one specified or master? 10:59
just curious
Geth rakudo/master: 5 commits pushed by (Elizabeth Mattijsen)++ 11:15
travis-ci Rakudo build failed. Elizabeth Mattijsen 'Don't close over anonymous classes in Uni' 11:38
travis-ci.org/rakudo/rakudo/builds/417860091 github.com/rakudo/rakudo/compare/4...6f512be718
buggable [travis build above] ☠ All failures are due to: failed make test (1 failure). Across all jobs, only t/04-nativecall/12-sizeof.t test file failed. 11:38
gfldex is there a maximum number of threads Rakudo is willing to spawn? 12:58
jnthn Depends precisely what you're asking. 12:59
You can call Thread.start({ }) until you run out of resources
Which then it's Rakudo's unwillingness :)
ThreadPoolScheduler, on the other hand, will do up to 64 by default
So if you have more than 64 cores and are CPU bound then a) I'm envious and b) you might want to increase it 13:00
MasterDuke oh, did we lose geth? 14:24
github.com/rakudo/rakudo/pull/2228 # any comments/thoughts would be appreciated 14:25
afk for a bit
AlexDaniel stmuk: it's what it says – 2018.06-449-g4f027ff63 vs 2018.06 17:06
AlexDaniel jnthn: by the way, there's tools/speedup.p6 tool that suggests unambiguous text descriptions for speed improvements 17:35
e.g. you say “so 1.95x faster than Rakudo” and the tool says 1.96x as fast
and yes it also round better :) 17:36
s
pmurias jnthn: yeah $_ being dynamic is super annoying to optimize, especially given that it is even declared when it's not used 17:44
AlexDaniel what about something like this:
m: for ^5 { say { $_ + 1 } … 5 }
camelia (1 2 3 4 5)
(2 3 4 5)
(3 4 5)
(4 5)
(5)
AlexDaniel is this behavior going to change? 17:45
stmuk AlexDaniel: I'm asking about the version of MoarVM not rakudo 17:49
AlexDaniel stmuk: ah, the one in NQP_REVISION and MOAR_REVISION 17:50
stmuk: but also I tend to bump before toasting, so it should be identical to master HEAD
timotimo AlexDaniel: that's just lexical, not dynamic
right?
stmuk AlexDaniel: thanks 17:51
pmurias does MoarVM use the truffle approach of being able to just assume stuff and when we see the assumption is actuall false being able to just throw away the assumption dependening code and deoptimizing? 17:57
AlexDaniel timotimo: yea I think you're right
pmurias so it would could a '$_ is never used as dynamic' assumption 17:58
timotimo yeah, that's the speculative optimizations we do 18:04
but just like the issue says, we'd have to throw away that particular assumption on any call to the outside 18:05
jnthn AlexDaniel: The usage you asked about is entirely lexical. A non-lexical one would involve CALLER:: or CALLERS:: or some such 18:15
AlexDaniel grep: CALLER 18:23
oh noes
AlexDaniel I should've waited 45 seconds… 18:24
[Tux] Rakudo version 2018.06-454-gfd6f512be - MoarVM version 2018.06-428-g727767006
csv-ip5xs0.930 - 0.944
csv-ip5xs-207.695 - 7.732
csv-parser24.957 - 25.949
csv-test-xs-200.441 - 0.466
test9.512 - 9.767
test-t2.172 - 2.191
test-t --race0.906 - 0.987
test-t-2035.408 - 38.169
test-t-20 --race12.544 - 12.681
AlexDaniel grep: CALLER 18:25
greppable6 AlexDaniel, 26 lines, 6 modules: gist.github.com/eba0e13e567dc66977...c3f945044e
AlexDaniel zef install Cro::HTTP::Router 19:21
===> Searching for: Cro::HTTP::Router
corrupted size vs. prev_size
Aborted
that's on HEAD
I tried it again and it seems to be working… 19:22
AlexDaniel where did this error come from though? 19:24
timotimo sounds like during gc it blew up 19:25
never seen that one
AlexDaniel heh 19:27
so p6bannerbot was kicked for excess flood because it attempted to give out too many +v ? :) 19:28
timotimo hm, possibly 19:29
AlexDaniel double free or corruption (!prev) 19:42
that's toast.6lang.org/module?module=Net:...it=2018.06
hm same on toast.6lang.org/module?module=Open...it=2018.06 19:44
but that's 2018.06?
so likely irrelevant
AlexDaniel hmm, something is not right, I'm just failing to make sense of it 19:48
so, here we see “corrupted size vs. prev_size” on 2018.06 toast.6lang.org/module?module=Pod:...it=2018.06
and I've *just* had it on HEAD 19:49
and I've never seen something like this previously with that frequency
so, does it mean that a change in zef is triggering some rakudo bug?
jnthn AlexDaniel: Dunno, but if it's a MoarVM regression there's not that many commits on HEAD vs. the last toaster run 20:06
I see lots of new regressions :/ 20:07
AlexDaniel jnthn: but how can it be a moar regression if it happens on 2018.06? 20:27
that's the part that is confusing
jnthn Oh 20:28
I didn't know that, I thought it was a regression since the last time we toasted, and explained the new set of problems
AlexDaniel yeah, but it was toasted by Zoffix
now that I did it, I had to toast 2018.06 freshly 20:29
so my completely uneducated and random guess is that it's some change in zef that is making moar blow up 20:30
*and* it's not something that was fixed on HEAD, because I just had it with freshly installed rakudo and zef on whateverable server 20:31
statisfiable6: help 20:32
statisfiable6 AlexDaniel, Available stats: core (CORE.setting size), install (size of the installation), libmoar (libmoar.so size) # See wiki for more examples: github.com/perl6/whateverable/wiki/Statisfiable
AlexDaniel statisfiable6: core
statisfiable6 AlexDaniel, OK! Working on it…
AlexDaniel, gist.github.com/ec5ee979637aab3197...dc74410df5
AlexDaniel statisfiable6: install 20:33
statisfiable6 AlexDaniel, OK! Working on it…
jnthn The latest zef commit is "Make ecosystem updates and listings multithreaded" :)
AlexDaniel oh
jnthn: yeah I didn't have time to even take a look, lol
:D
jnthn So I'm guessing that yes, it tickles a bug 20:34
statisfiable6 AlexDaniel, gist.github.com/91926c67a4b61a3557...e04029fa48 20:36
jnthn Did you get to do a toast of HEAD yet?
AlexDaniel jnthn: yes: github.com/rakudo/rakudo/issues/2230
the ones with ✗ and no comment are likely regressions 20:37
statisfiable6: libmoar 20:38
statisfiable6 AlexDaniel, OK! Working on it…
AlexDaniel or something, I'm still going through the list 20:39
jnthn OK, but does the zef thing seem to happen at HEAD too?
AlexDaniel yes! 20:40
in fact I had it happen right in front of my eyes when installing deps for whateverable, not just in toaster
jnthn ah, ok
statisfiable6 AlexDaniel, gist.github.com/3c96c478ce60e3898e...95de8acc21 20:41
AlexDaniel also, this `zef` change seems to rely on `race`, which IIRC is much more broken on older rakudos
so e.g. someone having early-2017 rakudo from distro repos will not be able to use zef properly, I think 20:42
jnthn Hm 20:43
And they'd not have a packaged zef?
I guess maybe not
AlexDaniel yeah, most distros don't have zef packaged
( github.com/perl6/user-experience/issues/29 ) 20:45
heh the commit message says “Speeds up `zef update` from 4.3s to 7.8s”
.oO( more seconds, more better )
jnthn Such is git...you write the commit message, then spot the typo after pushing :) 20:46
jnthn bbiab
AlexDaniel although… something about it does not add up 21:21
“ugexe committed 4 hours ago”
that commit wasn't even there when toaster was running
MasterDuke Zoffix: looks like Geth is dead 21:46
github.com/perl6/roast/commit/f6eb...526bd63b30 21:48
"Add some tests for applying offsets to Ranges"
AlexDaniel MasterDuke: pretty sure I can't ssh to hack 22:23
timotimo: ↑? 22:24
mmm this doesn't say what to do when you can't connect to it github.com/perl6/infrastructure-do...6c.org.pod 22:25
MasterDuke i guess some people have console access to the server, but i don't have any kind of account, so can't help at all 22:30
AlexDaniel so I'm looking at Audio::Sndfile right now 23:22
one of the test files segfaults semireliably
I just tested it on 2018.05 and there it doesn't 23:23
AlexDaniel so something is going on and I should check it 23:23
one problem is that the module has 12 dependencies and most of them are not in perl6-all-modules :S
is there any easy way to tell zef to install a bunch of modules into a folder? 23:24
so that I can then have a single location to set PERL6LIB to?
jnthn I think there's some --to=
(But don't know the details, sorry) 23:25
AlexDaniel ok I tried `zef install --install-to=. JSON::Fast` and it created a bunch of folders with stuff 23:26
that's nice, but then?
jnthn Can you then put . in PERL6LIB and have it work? 23:28
AlexDaniel tried that, doesn't work
jnthn Hmm
What if it's --to=file#. if it allows that?
AlexDaniel no, I don't think so 23:31
but is there no way to point perl6 executable to a whole repository?
jnthn Does -Iinst#. or some such work? 23:35
iiuc, it's kind-of-repo#path
AlexDaniel jnthn: Yes!!!
jnthn :) 23:36
AlexDaniel ok that is awesome
I should've figured it out earlier, will make my work so much easier
or, well, asked :)
MasterDuke jnthn: you had some comments yesterday re optimizing a Range with an offset, any thoughts on whether github.com/rakudo/rakudo/pull/2228 is the right approach? 23:43
timotimo i was just able to reach hack via ssh 23:52
AlexDaniel timotimo: cool, how is it?
timotimo ... 23:53
timotimo ok i can reach it but not execute htop 23:53
yeah, i bet it's hanging
AlexDaniel you can maybe issue a reboot?
timotimo i am about to