AlexDaniel` Yeey :) 00:46
MasterDuke .tell brrt i added the loads to the sp_findmeth template and updated the PR. rakudo builds ok, any easy way to be sure that path is being taken? 00:49
yoleaux MasterDuke: I'll pass your message to brrt.
Zoffix samcv++ 00:50
MasterDuke .tell brrt running into some weirdness with a box_i template, some details/discussion here irclog.perlgeek.de/moarvm/2018-01-24#i_15733736
yoleaux MasterDuke: I'll pass your message to brrt.
samcv anybody getting failures on t/spec/S32-io/IO-Socket-Async.t 00:59
i get this if i run it without prove: Unhandled exception in code scheduled on thread 01:00
4
MasterDuke i've run a couple spectests today, haven't seen that
Zoffix samcv: try changing the $port in the test? Maybe you got something blocking it 01:04
samcv ok
Zoffix: no change 01:05
gonna try compiling withou --has-ffi
nope still failing :( 01:06
Zoffix :(
samcv ah i think i have ipv6 disabled?
ok that makes sense then :)
MasterDuke ah, right. i got that on my desktop all the time when i had ipv6 disabled also 01:07
samcv though it having unhandled exceptions doesn't sound great either :|
MasterDuke i think the test is making an assumption good user code wouldn't 01:08
Geth MoarVM: 67cde5bf03 | (Samantha McVey)++ | VERSION
Bump VERSION to 2018.01
 01:11
samcv tag pushed 01:12
released! 01:15
Zoffix woooo \o/ samcv++ 01:18
TimToady also sez samcv++ 01:20
samcv MasterDuke: oops missed your comment 01:22
AlexDaniel samcv++ indeed! 01:39
MasterDuke samcv++ i don't think a stray . will break the release 01:44
AlexDaniel perhaps time to start calling it Chaneglog, accepting that it may be less than awesome :) 01:54
timotimo should the "feel free to edit this page" have stayed in the changelog? :S 01:58
AlexDaniel oh, oops 02:00
and “New in 2018.01” twice 02:01
timotimo i find it slightly strange that there's only a single entry that has an explanation attached to it (the last one in the spesh section) 02:02
Geth MoarVM: 726ca9b596 | (Aleks-Daniel Jakimenko-Aleksejev)++ | docs/ChangeLog
Remove unneeded text

It was moved from the wiki and shouldn't be in the actual changelog.		 02:03
TimToady in the fwiw department, I experimented with a config.o that had extra null padding on all the strings containing the PREFIX directory, and it seemed to work fine, so I suspect making a relocatable binary won't be all that difficult 02:12
TimToady 'course, things that copy PREFIX around inside nqp may have to be a bit more dynamic in adding padding 02:16
timotimo TimToady: could probably also use ./././././././ as i think samcv did in the appimage stuff 02:18
TimToady yeah, thought about that, but it's uglificatious and has performance implications :) 02:19
timotimo fair enough
TimToady hopefully we can keep files distinct that need padding vs those that can just s/// the text 02:20
and since each ./ is presumably a system call, it might get worse with recent CPU security patches 02:23
though arguably most such paths aren't used in a hot path
also, arguably, one could use the relocator to remove the ././././ spacers just as easily from the standard install, as long as the relocator knows which ones to null pad, and which ones to s/// 02:25
timotimo we already do a crapton of superfluous syscalls on startup; many times we stat the same thing many times in a row
TimToady (and it tends to be obvious by whether the string is \0 terminated already) 02:26
timotimo perl6 -e '' stats ModuleLoader.moarvm 81 times for example
more precisely, almost all of those are lstat
TimToady nice to still have optimization targets
Zoffix :) 02:28
timotimo almost all of these come from MVM_file_exists 02:36
gist.github.com/timo/33acff43fa16a...e98705a832 - go nuts 02:38
fwiw, i have not actually timed if stat takes any meaningful amount of time out of our startup
callgrind seems to think it's only a tiny amount of instructions spent, but i'm not sure how it measures syscalls, if at all. 02:40
brrt good * #moarvm 07:17
yoleaux 00:49Z <MasterDuke> brrt: i added the loads to the sp_findmeth template and updated the PR. rakudo builds ok, any easy way to be sure that path is being taken?
00:50Z <MasterDuke> brrt: running into some weirdness with a box_i template, some details/discussion here irclog.perlgeek.de/moarvm/2018-01-24#i_15733736
brrt .tell MasterDuke seeing it, I will check it out as well. I was hoping you'd find bugs like these :-)
yoleaux brrt: I'll pass your message to MasterDuke.
brrt also, that looks sane, it should probably just work 07:18
samcv i'm curious does ././././ actually use more cpu or does the system just optimize it away? 07:19
brrt .tell jnthn that $0 always refers to the address rather than the register value
yoleaux brrt: I'll pass your message to jnthn.
brrt ans all others do refer to the register value 07:20
samcv: what's the context?
samcv brrt: any
tbh
brrt eh. and this is file paths? 07:21
i suppose the system can optimize it away. i'd be very surprised if you could measure that
nine Fun fact: the inline in place stuff actually reduces the code size: 5 files changed, 100 insertions(+), 160 deletions(-) 07:43
brrt \o/ 07:48
samcv nice! 07:49
Geth MoarVM/master: 16 commits pushed by (Stefan Seifert)++, (Timo Paulssen)++
review: github.com/MoarVM/MoarVM/compare/7...39928851f8		 08:01
nine
.oO(pushing largish branches is just not as satisfying)
samcv++ # release 08:04
travis-ci MoarVM build failed. Samantha McVey 'Update ChangeLog for release' 08:57
travis-ci.org/MoarVM/MoarVM/builds/333049969 github.com/MoarVM/MoarVM/compare/c...b09504eeb3
lizmat Thanks everybody for the MoarVM release! 09:06
moarvm.org # still showing 2017.09 :-( 09:07
which could give people the wrong idea that there's no development with MoarVM anymore
samcv i thought that was fixed? 09:40
i'm not really sure how the site works :\
oh the releases page shows 2017.12 09:41
i think we need an update to the release document to make sure the rest of the site is updated properly? jnthn
jnthn Yes, we can put that into the release document 10:09
yoleaux 07:19Z <brrt> jnthn: that $0 always refers to the address rather than the register value
jnthn github.com/MoarVM/moarvm.org is the git repo for the site, and it auto-deploys
Anybody who has time/interest to help keep it up to date, please just ask for a commit bit :-)
(I've got interest, of course, but apparently not time...) 10:10
travis-ci MoarVM build failed. Samantha McVey 'A few ChangeLog adjustments' 10:29
travis-ci.org/MoarVM/MoarVM/builds/333050302 github.com/MoarVM/MoarVM/compare/7...22511e3a51
El_Che samcv, jnthn: It seems the new 2018.01 release is missing from moarvm.org/releases/. Can you upload a tar? I needed to build the rakudo-pkg Linux OS packages for Rakudo 11:29
thx
jnthn El_Che: Seems it wasn't committed: github.com/MoarVM/moarvm.org/commits/master 11:33
samcv: ^^
Or at least, not pushed 11:38
timotimo --collect-systime=<no|yes> [default: no] 13:00
This specifies whether information for system call times should be collected.
it estimates a "sysTime" of 1 for MVM_file_exists, but for some reason MVM_repr_set_int has a self sysTime of 2.599?? 13:02
travis-ci MoarVM build passed. Aleks-Daniel Jakimenko-Aleksejev 'Remove unneeded text 13:09
travis-ci.org/MoarVM/MoarVM/builds/333083190 github.com/MoarVM/MoarVM/compare/6...6ca9b596ea
jnthn timotimo: Anything that allocates can trigger GC. A top allocator will thus show some of the GC cost 13:09
timotimo ah, i suppose we maybe do a single gc run during startup - though i had believed we were down to 0? 13:10
jnthn Hmm 13:12
If it's startup then no idea
timotimo yeah, i was measuring perl6 -e '' 13:13
i was hoping i could get more details on how costly the lstats are
jnthn ah
timotimo i expect strace to be rather overhead-heavy. maybe i'll try perf trace, that's allegedly a lot lighter 13:17
65.483 (18446744073709.523 ms): moar/26304 mprotect(start: 0x7f3860279000, len: 8192, prot: READ|WRITE ) ... 13:18
i've seen this somewhere before ...
timotimo adding all lstats and stats up gives me 0.54 miliseconds 13:21
lizmat wow
ah, 0.54 *milli* seconds
that's not too bad then, is it ? 13:22
timotimo with some care we could probably cut it down to a third of that?
not terribly bad, no
fwiw, all our mprotect calls added up give me 24.8 milliseconds ;) 13:25
lizmat now, *that* I call significant
timotimo without jit it's only 0.69, but that makes me thing all of that is on the background thread 13:26
brrt imho 24.8 milliseconds is a reasonable price to pay to have some security against (some) attacks 14:36
nine brrt: it's only on the spesh worker thread anyway, isn't it? 14:39
brrt yep
timotimo the stats, on the other hand, are all on the main thread ;) 14:41
jnthn Stats?
Oh, recording of stuff
timotimo lstat and stat
jnthn Oh, right, not spesh stats :D
haha
timotimo yeah :)
psch ooc, is moar on a SPARC T2 interesting? a friend has access to a T5140 iirc 16:22
japhb psch: Well, it won't have JIT, but porting to it might expose some remaining portability bugs in the main codebase. 16:23
psch it's running debian currently, although there was talk of throwing one of the free solaris forks on it, if that matters 16:25
but then i also don't know how liberally they want to hand out shells
timotimo is that a LE or BE architecture?
i *think* we're currently not doing little/big endian conversion properly everywhere
psch so at worst this might be nothing more than "well i tried building it here's the errors: ..."
geekosaur SPARC is BE 16:26
ilmari _and_ alignment-picky 16:27
japhb Which makes it 1) great for weeding out portability bugs, and 2) probably a LOT of work 16:28
ilmari ubsan already catches unaligned reads 16:33
Configure.pl (well build/probe.pm) disables CAN_UNALIGNED_* under -fsanitize=undefined
stmuk Is the moarvm release tarball likely to appear at moarvm.org/releases/ ? 17:54
El_Che stmuk: i pinged jnthn and samcv. It should soon proably 17:55
stmuk ah great
AlexDaniel I guesss we'll need to wait for samcv for that, yea 19:00
oh 19:01
I see it's there already?
stmuk, El_Che ↑
stmuk ah great!
El_Che restarting the test build: travis-ci.org/nxadm/rakudo-pkg 19:04
AlexDaniel samcv: fwiw the main page on www.moarvm.org/ talks about 2017.09 19:37
samcv jnthn: pls help with site
do we have a way to generate the site automatically?
AlexDaniel also www.moarvm.org/releases.html has no 2018.01 19:38
El_Che everything is there now
my builds pass
AlexDaniel yey
El_Che (running, but most of them passed already)
samcv we need a script to generate the page i guess? 19:39
the tarball you can dl it's just not linked..
El_Che thx samcv 19:40
just added opensuse support 19:42
lizmat El_Che++
samcv maybe i'll come up with a script if nobody has one 19:43
jnthn samcv: No script to update it as of yet, but that'd probably be wise 19:52
samcv ok i will write one!
working on it right now
jnthn Cool
I've kinda meant to for ages, but something always pushed it off the top of the priority list 19:53
samcv seems important since the site keeps falling behind
jnthn *nod*
El_Che thx samcv, building the pkgs now 20:04
samcv almost got it i thkn 20:28
Geth MoarVM: f1197e42ed | (Samantha McVey)++ | docs/ChangeLog
Edit the ChangeLog to include the original release

Also fix the top heading. This was added so we can generate
  moarvm.org/releases.html from this document easily and this release was
missing from the ChangeLog file but existed on the website.		 23:55

Powered by Cro and the Raku® Programming Language.

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

Thank you!