github.com/moarvm/moarvm | IRC logs at colabti.org/irclogger/irclogger_logs/moarvm
Set by AlexDaniel on 12 June 2018.
Geth MoarVM: 9d4681afe6 | (Daniel Green)++ | 7 files
Rename getstrfromname to strfromname

This better fits the convention of the other `*from*` ops.		 03:20
MoarVM: 9b44d4816d | (Daniel Green)++ | src/jit/graph.c
JIT strfromname in the legacy JIT
 03:21
MoarVM: 46edb6f63c | MasterDuke17++ (committed using GitHub Web editor) | 8 files
Merge pull request #899 from MasterDuke17/rename_getstrfromname

Rename (and JIT) getstrfromname
Zoffix New blog post: "Introducing: Perl 6 Marketing Assets Web App": rakudo.party/post/Introducing-Perl...ts-Web-App 03:42
jnthn morning o/ 09:26
masak \o 09:43
brrt \o 11:04
lizmat brrt: o/ 11:05
brrt ohai lizmat 11:06
3 more days, eh
jnthn 4 including today for me, I think... 11:18
brrt :-(
heh, the spesh thread, when stopped, wants to run a VM frame.... 11:21
because the spesh_worker is just it's initial_invoke()
jnthn The presumption was it'd never be stopped. :) The event loop thread is the same. 12:14
dogbert2 jnthn: added the 'false positive' bug from yesterday: github.com/MoarVM/MoarVM/issues/930 13:16
timotimo i.imgur.com/UI4mZE6.gifv - jnthn, did you see this? :) 13:21
lizmat
.oO( it seems to go on forever :-)
 13:25
jnthn timotimo: yes :-) 13:32
Needs a bit more explanation in the UI of the numbers I guess, but otherwise...neat :D
brrt notes that it would be very, very easy to implement something like this for MoarVM: github.com/dinfuehr/dora/blob/mast...os/perf.rs 13:36
nwc10 jnthn: ./perl6-m -Ilib t/spec/S02-lexical-conventions/unicode-whitespace.t 14:02
src/core/interp.c:2814:17: runtime error: member access within null pointer of tfc006caf6e1 sp 0x7fff273dee80 bp 0x7fff273e1020 T0)
#0 0x7fc006caf6e0 in MVM_interp_run src/core/interp.c:2814
#1 0x7fc0071a3302 in MVM_vm_run_file src/moar.c:412
and ASAN still very excited by t/spec/S29-conversions/ord_and_chr.rakudo.moar 14:03
I torment my bild with 14:04
+#define MVM_ARRAY_CONC_DEBUG 1
+#define FSA_SIZE_DEBUG 1
+#define MVM_SPESH_CHECK_DU 1
MVM_SPESH_BLOCKING=1
MVM_SPESH_NODELAY=1
no idea if it hides without any/all of these set
dogbert2 now that's torment :-) 14:07
brrt oh, and we're destroying the FSA twice in --full-cleanup mode 14:09
dogbert2 interesting, --full-cleanup has been busted for a loong time 14:11
dogbert2 the ASAN excitement posted by nwc10++ points to github.com/MoarVM/MoarVM/blob/mast...ker.c#L331 14:13
dogbert2 must be run with MVM_SPESH_NODELAY=1 and MVM_SPESH_BLOCKING=1 in order to appear 14:14
dogbert2 nwc10: github.com/MoarVM/MoarVM/issues/931 14:19
timotimo jnthn: i'm planning to pepper the whole UI with so many explanations that it'll be easy to understand even for newcomers ... or so is the hope 15:06
timotimo i'll have to find a sensible approach to pagination or maybe just a "load more" button would be enough 15:34
lizmat but where would the "load more" come from? From a server, or would it already live in the browser ? 15:35
timotimo there's a server that can do that 15:36
diakopter dogbert2: in that code, I think lexreg can move since sf can move, so the lexreg pointer chasing must be done inside the loop 15:37
nwc10: I bet if you re-run it with moving that lexreg assignment inside the loop it'll work 15:40
lizmat timotimo: then a "load more" button would be appropriate :-)
timotimo: but otoh it would be nice if you could send such a profile to a gist e.g. 15:41
timotimo: so how would that work ?
timotimo well, the server could interface with github, for example 15:43
or do you mean the whole profile should go on github?
Geth MoarVM: aa49992398 | (Matthew Wilson)++ (committed using GitHub Web editor) | src/spesh/frame_walker.c
rechase lexreg each time through the loop

let's see if it fixes github.com/MoarVM/MoarVM/issues/931		 16:00
diakopter nwc10: ^^
apparently github thought "let's see if it fixes" means "fixes" 16:03
lizmat timotimo: I was thinking of the case when people see something weird in a profile they did and they want to show it to other people 16:11
until now, that would have been posting the HTML file as a gist somewhere 16:12
timotimo true 16:14
now it'll be an .sql file instead, or maybe the .sqlite database file
lizmat ah, that would work, I guess 16:15
timotimo though i assume it'd also be nice to put summaries up as gists, too 16:17
dogbert11 diakopter: unfortunately it seems as if the problem is still present. The stacktrace is the same as before. 16:26
timotimo twitter.com/python_tip/status/1026...5891014656 - surely this is a thing perl6 gets right 16:30
lizmat m: dd "Welcome in 🇬🇧".flip # indeed :-) samcv++ 16:38
camelia "🇬🇧 ni emocleW"
lizmat could someone reply to that tweet with the Perl 6 version ^^ ? 16:39
timotimo i won't make a screenshot of that because in my terminal it shows as two overlapping(!) letters in boxes %) 16:40
lizmat timotimo: perhaps refer to colabti.org/irclogger/irclogger_log...08-06#l766 ? 16:42
timotimo in my browser it's a circle with black-and-white lineart of the union jack in it 16:43
that's not so much better
lizmat timotimo: I mean, use that link, not a screenshot of the link 16:49
timotimo hm, how about a link to a "run it online" service including a screenshot of the result 16:50
lizmat well, that won't work for you either then
your browser apparently has problem displaying 🇬🇧 16:51
timotimo right
someone else could do it, though
lizmat so you do want to put a screenshot there instead of a link to the log ? 16:55
or do you think you cannot actually put the right unicode in the tweet ?
timotimo i can put the right unicode in there, i believe 16:57
i dunno
lizmat doesn't do Twitter 17:01
timotimo a wise decision
lizmat www.reddit.com/r/ProgrammingLangua..._floating/ # did anybody look at this ?
timotimo doesn't do facebook
lizmat neither 17:02
timotimo i believe zoffix saw it
could be a different one?
jnthn heh, the recent Comma interview I did made it to the Facebook page of JetBrains (the company behind the IntelliJ platform) and of course some folks just had to trot our their tired Perl memes. 17:05
Zoffix Yeah, it's the same algo someone pointed out last week. The URL I have is: pldi18.sigplan.org/event/pldi-2018...conversion 17:06
jnthn Zoffix: Is that the one you implemented a while back? 17:08
Zoffix Filed as R#2187 so we don't forget
synopsebot R#2187 [open]: github.com/rakudo/rakudo/issues/2187 Look into implementing faster f.p. stringification algo
Zoffix jnthn: no, I did Dragon4. This one is supposedly better
The person who showed me the video said the guy did compare the new algo to dragon4 in the video (in pldi18.sigplan.org/event/pldi-2018...conversion ) but I didn't have time to watch it yet 17:09
jnthn Ah, ok :)
jnthn looks through the PLDI program to see if there's any papers he should read :)
Zoffix oh wait, not Dragon4...
Grisu3 :)
Grisu3 with Dragon4 fallback was state of the art before that new algo, and we use Grisu3 with spritnf fallback 17:10
*sprintf
timotimo i wonder if it'd be faster to fall back to dtoa? as it doesn't have to "parse" the "%f" or whatever we pass 17:11
lizmat jnthn: do you have a link to that mention on the JetBrains Facebook page ? 17:15
stmuk www.facebook.com/JetBrains/posts/1...1294013449 17:17
we need someone to defend the beard as well :) 17:18
timotimo i think it was meant as a compliment, though?
on the beard, i mean
stmuk probably! 17:19
timotimo personally i like jnthn's beard 17:20
Zoffix Yeah, I read it as a compliment too...
diakopter jnthn: the PLDI keynote video by Margaret Martonosi is worth watching (takes a few minutes to get meaty): youtu.be/VwNkMioYRoU 17:33
pldi18.sigplan.org/event/pldi-2018...onosi-talk 17:34
or did you mean a different PLDI
timotimo that sounds like the right pldi 17:36
diakopter dogbert11: oh noes, that means the rooting error is in one of the things invoked within that block 17:39
jnthn Going Facebook and Twitter engagement, "jnthn's beard" would be a successful Perl 6 marketing alias :P 17:40
*going on
diakopter: That looks kinda interesting indeed :) 17:41
The OSR one looks quite relevant :)
timotimo jnthn: that could be a band, you know
jnthn Were I called Spock it would be already :P
diakopter jnthn: I was in attendance for the final day, since I now live in a suburb of east Philadelphia 17:42
jnthn ah, cool :)
timotimo yup
i haven't actually listened to spock's beard yet 17:43
brrt seems we're still leaking like a watering can from the spesh thread, even after we shut it down :-) 18:24
jnthn I think it keeps various bits of state that might need clearing up 18:25
brrt oh well, we'll fix it someday 18:27
timotimo you got any interesting results to share? 18:34
maybe i can find one or two "free"s to place
[Coke] BY JNTHN'S BEARD THIS MUST BE OPTIMIZED 18:50
dogbert11 diakopter: I doubled the following line github.com/MoarVM/MoarVM/blob/mast...ker.c#L336 with a 'MVMuint16 type2 = ...' and the number of valgrind complaints doubled 18:56
diakopter dogbert11: interesting :) now you're making me suspect MVMROOT3 18:58
it would really suck to have to put find_lex_info inside the loop too 18:59
timotimo m: my @a; say @a[0..100000] 19:00
camelia ((Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any) (Any…
timotimo i was looking at the old commit again %) 19:01
diakopter wait, does this look right MVMObject *obj = frame->env[idx].o; 19:07
actually, should this be checking for vmnull?? if (!obj) (right after the above line) 19:14
timotimo we sometimes use Actual Null to signal something, i think? 19:16
diakopter timotimo: surely someone could write a static analysis tool to recursively scan MoarVM source code, marking routines that could possibly allocate/GC (and all their callers), so that all the missing MVMROOT invocations could be found
timotimo: or anyone I mean
timotimo clang has a framework that lets you write analysises like that 19:20
diakopter actually I think it's accessing the hll object unsafely 19:27
wait no, DUH, I see the problem
Geth MoarVM: d3b11dd3e9 | (Matthew Wilson)++ (committed using GitHub Web editor) | src/spesh/frame_walker.c
root when allocating
 19:30
diakopter dogbert11: can you try that
timotimo that doesn't look right 19:31
at the point ctx_hash is allocated, sf and frame are garbage pointers
uninitialized
they only get initialized by find_lex_info
diakopter oh right
hrm 19:33
ehhh how to rollback that commit :D
I mean, it shouldn't break anything, but, pessimization 19:34
er, actually it would break.
oops.
can someone revert it? 19:35
or I can in the editor
timotimo it can break, yeah 19:37
Geth MoarVM: eee961e0a8 | (Matthew Wilson)++ (committed using GitHub Web editor) | src/spesh/frame_walker.c
messily revert the last
samcv good * lizmat 19:38
lizmat samcv o/
diakopter omg messed it up again 19:41
Geth MoarVM: 5bfd2c748a | (Matthew Wilson)++ (committed using GitHub Web editor) | src/spesh/frame_walker.c
:(
brrt diakopter: I think the bug you wanted to fix was the correct one, but please use git proper rather than messing with the web editor, especially on master; you can't compile your edits that way 21:30
diakopter brrt: yeah 21:31
brrt it's no big deal :-)
diakopter I think it's not the bug; rather I suspect it's something with how the lexical slots are accessed
(that happens to not error most of the time) 21:32
brrt well, allocating without rooting certainly could be a bug
what bug are we talking about?
diakopter right, but as timotimo points out, those pointers are given new values in the next line anyway
(and they're unitialized before that)
brrt I like the idea of static analysis (although I'm afraid we're using a bit much in the way of macros for that to be performant :-)) 21:33
timotimo the right solution to this conundrum, with web editor and such, is obviously to compile gcc to javascript using emscripten so you can compile moarvm right there in your browser 21:40
brrt obviously
timotimo our most valuable developers all have something in common: 21:41
they own a smart phone
well, i think most of them do
brrt not one that can compile moarvm :-) 21:47
although it isn't all that much
timotimo just connect an usb stick via usb OTG and have a swap file on it 21:48
or perhaps store the swap on a sufficiently big NFC tag
"please touch nfc tag number 7 to the phone now"
2013 someone said "NFC tags can store upwards of 8kB of data depending on the IC, but the larger memory payloads can get expensive." 21:49
jnthn I suspect mine is easily capable, hardware wise, of compiling MoarVM. It's mercifully far too locked down for me to do any such fun. :)
timotimo what kind do you use? 21:50
jnthn iPhone 21:51
timotimo it almost sounds like the protocol in use for storing and retrieving data with NFC tags won't let you address more than 8 kbytes?
yeah, those are locked pretty tight
jnthn Which has worked for a bunch of years without problems, and does all I care for.
timotimo mine recently got a crack in its display 21:52
i'm considering upgrading, or maybe "sidegrading"
i.e. getting a cheap smartphone that's still a bit more powerful than my current one
jnthn My wife has some Android-based phone and it's a security shitshow. Claims its running the latest available version (I presume this means "that the vendor makes available"...) 21:53
timotimo yeah, vendors can be pretty bad about that
jnthn But then gets itself infected within a week or two with poker games and who knows what else :/
timotimo for many devices there's community-managed custom roms
but .. mehhh
i have not yet had a security problem with any of my android phones, but i'm just a little paranoid 21:54
jnthn Well yeah, for the most part I just want my smartphone to do its routine/boring things and do so reliably. 21:55
timotimo as in i refrain from doing stuff normal users would probably want to be able to do, but that might be risky in some way
jnthn Not have to mess around fixing it 21:56
timotimo i'd actually be interested in playing the odd game every now and then
right
i mean i could also schlepp my nintendo 3ds around with me and get into homebrew development ... there isn't really amazing homebrew games for that out yet
where saying "yet" makes it sounds like "it's just because the system hasn't had homebrew for long", but some people are already calling the system dead %) 21:57
jnthn I used to carry around a monocular so that I could read information boards at public transport locations and road name signs placed uncomfortably high and so on (since I've got pretty awful eyesight). Now I just carry the smartphone and see all the information on that. So I rather care that it works, because I could be quite literally lost without it. :-) 21:58
timotimo rightfully so 21:59
i believe i've seen a service out there, an app plus volunteers, that lets a volunteer "be your eyes" through your phone's back-facing camera
accessibility and assistive technologies is quite an interesting space 22:00
diakopter jnthn: I carry a fully-charged backup phone since dropping my phone once left me stranded for hours :O
timotimo oh damn
that's an experience i'd not like to have
jnthn Yeah, it seems the kind of tech that smartphones and tablets have made cheap and portable is quite useful for assistive stuff, which is a nice side-effect. :)
I also have nystagmus, which isn't a very big deal, but does mean that my eyes don't fixate. Went to teach at a company that does eye tracking and they were saying "oh, and it's good for lots of accessibility things", then none of their demos would work on me at all. :D 22:03
timotimo oooh, i haven't heard of that yet
lizmat and another Perl 6 Weekly hits the Net: p6weekly.wordpress.com/2018/08/06/...thon-done/ 22:04
jnthn Another bug down 22:05
oh, not this window :)
diakopter bug++ 22:07
timotimo BBL ~ 22:08
Geth MoarVM: ac3d3c7695 | (Tobias Leich)++ | src/6model/reprs/CStruct.c
Fix alignment/mem issue of inlined CArrays

Now CArrays that are inlined in CStructs are handled correctly regarding alignment of the CArray within the struct. Also accessing the slots of the inlined array works.		 22:20

Powered by Cro and the Raku® Programming Language.

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

Thank you!