Perl 6 - the future is here, just unevenly distributed

IRC log for #moarvm, 2017-04-04

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary

All times shown according to UTC.

Time Nick Message
00:06 timotimo looks like incorrect fudging?
00:07 timotimo i.e. it reaches the end, but it doesn't reach the specified amount of tests
00:07 timotimo (file-ops.t)
00:59 samcv yay appveyor
00:59 samcv can we install clang on windows
01:00 samcv well of course we can. will be cool to do coverage there too.
01:00 samcv when i'm feeling really masochistic maybe i'll do that
01:17 samcv "Perl: As Perl developers continue to wait for the arrival of Perl 6, other languages have seized the spotlight. Though some developers still find this language efficient and portable, it’s not without its annoyances. The popularity of Perl is fading fast and, unless Perl 6 is introduced soon, the demand for this technology may die out."
01:17 samcv oh my
01:23 [Coke] .seen samcv
01:23 yoleaux2 I saw samcv 01:17Z in #moarvm: <samcv> oh my
01:23 samcv hi
01:23 [Coke] pmsged
02:20 geekosaur joined #moarvm
02:26 evalable6_ joined #moarvm
02:59 evalable6 joined #moarvm
04:14 evalable6 joined #moarvm
05:21 stmuk joined #moarvm
05:53 domidumont joined #moarvm
05:59 domidumont joined #moarvm
06:05 brrt joined #moarvm
06:42 brrt joined #moarvm
06:52 brrt timotimo: the other solution is to install gnu make or bsd make :-P
06:52 brrt or mingw
07:54 zakharyas joined #moarvm
08:12 samcv oh guys
08:13 samcv i have roast coverage reports now too
08:52 brrt hi samcv
08:52 samcv if you guys curious
08:53 samcv Roast: line based (per file) https://cry.nu/coverage/roast/
08:53 samcv function and file summaries of coverage % https://cry.nu/coverage/roast.html
08:53 samcv really nice, you can see which code doesn't get executed which can sometimes be insightful
08:54 samcv meaning not just code that is not tested, but code that may never execute or maybe was created with conditions in mind that may not actually happen
08:56 brrt huh, that's pretty funny
08:56 samcv funny?
08:57 samcv also for #define's it will show the number of times code USING that thing has been run
08:57 samcv which is pretty mindblowing. but
08:57 samcv i verified it
08:59 dogbert17 joined #moarvm
09:13 samcv brrt, maybe you will find something interesting
09:13 brrt funny as in interesting :-)
09:14 samcv brrt, https://cry.nu/coverage/roast/coverage/home/samantha/git/MoarVM/src/jit/log.c.html
09:14 samcv maybe should inline that
09:14 samcv 69.1M invocations. one of the highest i've seen
09:16 brrt hmm
09:16 brrt the jit logging needs to be cleaned up anyway
09:16 samcv interesting here https://cry.nu/coverage/roast/coverage/home/samantha/git/MoarVM/src/jit/compile.h.html
09:16 samcv see how it shows the number of uses of that one define
09:17 samcv and the other one has no uses. and i used a grep and it is actually used never
09:19 samcv found (refound?) some functions in normalize.c that used to be used to determine regional codepoints and spacing marks but no longer are needed. so prolly gonna remove the spacing mark one. since it doesn't have much use. and the code inside it is sub-optimial
09:21 samcv err that's in anotherfile. but regardless. heh. found some code in normalize.c that i thought never would get executed/never understood why it was there. and turns out it doesn't execute lol
09:21 samcv or maybe it is needed, idk. just i guess we've never used the function in that situation
09:34 timotimo at first i thought "roast line based per file" meant you had one whole coverage report for every roastfile
09:41 timotimo that'd be a firehose of information; i'd expect we'd have to bucket it for it to make sense
09:41 timotimo like, roast file 01-foo/bar.t covers 10 files at 0%, 7 files at up to 10%, 9 files at up to 20%, ...
09:42 samcv i do timotimo
09:42 samcv but it's not online
09:42 timotimo oh
09:42 samcv i can make it if you want XD
09:42 timotimo is it many gigabytes big? :)
09:42 samcv idk
09:42 timotimo i'm not sure if it'd actually be helpful
09:43 samcv when roast runs it generates them in a directory format similar to roast itself
09:43 dalek joined #moarvm
09:43 samcv but they are in coverage/t/spec/S05-blah/mything.t.covraw (or whatever the extension is i forget)
09:43 samcv then i merge all of them together in the t/spec folder (so i only get profiles for roast files)
09:44 samcv it takes the 1st command line argument passed to perl6-m so i have lots of weirdly named ones not in any folder
09:44 samcv --help.profraw -IGNOREME.profraw lol
09:44 timotimo ugh :)
09:45 samcv but all the roast ones are nicely in folders exactly how they are in roast
09:45 samcv just ignore the files in ./coverage and only check t/'s files and it's fine
09:45 samcv ThisDoesNotExistAtAll.p6.profraw
09:45 samcv lol.
09:46 samcv i guess that should maybe be included? but that will require some deciding how to do it
09:46 samcv not sure
09:47 samcv well they're 535MB
09:47 samcv the profile files for all the roast tests XD
09:47 samcv they are 416K when processed though
09:48 samcv same as they are always. since it doesn't need much more space to record it running some section of code 100000 times compared to 1 time
09:48 timotimo yeah, i don't think it'd make too much sense to upload the raw ones
09:48 samcv since it's all dependent on the size of MVM's codebase
09:48 samcv yeah, obv
09:48 timotimo right
09:48 samcv but i may eventually allow you to check coverage of specific roast files.
09:48 samcv i mean why not right. go all out
09:48 timotimo these reports should compress really well when you put the same files next to each other in a compressed archive
09:51 samcv gonna try compressing the raw reports :P
09:51 samcv by default it's default.profraw, and file gets replaced each time it's run. i'm not sure if that actually gets added to the old one or not. but it's better to be able to seperate out the test files
09:52 samcv and only merge the ones you know are actually tests, and not like, compiling the compiler
09:52 timotimo do you think it'd be worth trying to have clang on windows? like, we already have that on travis
09:52 timotimo but what we could try is mingw? but i think it's known to not work, at least for the rest of rakudo?
09:52 samcv idk how much effort is it. it must be a normal thing for appveyor? maybe has a tutorial on it?
09:53 samcv well the 500MB raw files compressed to 22MB with XZ compression
09:53 samcv not bad. and showed off my cpu :)
09:53 samcv since xz can be threaded which is neat
09:54 timotimo oh, neat indeed
09:54 samcv we def need better coverage of NativeCall
09:54 samcv hopefully we can use this info to add more tests
09:55 samcv also not sure diff between cover and executed https://cry.nu/coverage/roast.html
09:56 samcv but i think must have to do with possibilities maybe? compared to just line by line
09:56 timotimo perhaps it's how many functions out of this file were ever entered?
09:56 samcv because some lines can execute different ways?
09:56 samcv nah
09:56 samcv could be the regions?
09:56 timotimo that'd be path coverage, then
09:56 samcv instead of lines?
09:56 samcv yeah that's what i *think* it is
09:56 timotimo you mean if one line has multiple coverable pieces?
09:56 samcv yep
09:56 timotimo like if (foo) { puts("lol") } else { puts("nope") }?
09:56 timotimo that'd be three individual pieces
09:57 samcv sure why not lol
09:57 samcv may be easier to see here on nqp tests: https://cry.nu/coverage/report-libmoar.html
09:57 samcv since it shows cover vs live coverage
09:58 samcv for functions
10:03 timotimo i'd dig it if we could highlight the path parts after MoarVM/src/
10:06 samcv yeah
10:06 timotimo should be easy with sed, but there doesn't seem to be an ansi class for like a bright white
10:06 samcv been thinking about building in in /tmp just so it has a shorter path
10:06 samcv it's called bold
10:06 timotimo oh, ok
10:06 samcv ;)
10:08 samcv anyway lovely people it's about my bedtime
10:08 samcv i think i figured out how to fix this page https://cry.nu/coverage/report-moar.html
10:09 timotimo why not just throw out the coverage report for the moar binary?
10:09 samcv to not be a graveyard of files not in that file. and also make sure i don't miss .h files with code in them. will just process with head/tail and cut the output from the file based coverage report and pass it in
10:09 samcv why should we
10:09 samcv i'll probably appand it to the end of the other one
10:10 samcv but in case there later become more files that are used in that binary it could be useful
10:10 samcv or maybe there's a way to combine libmoar and moar.... but idk
10:10 jnthn samcv++ # just glancing the coverage reports, looks pretty great :)
10:10 samcv thanks :)
10:11 timotimo OK
10:11 zakharyas joined #moarvm
10:22 debugsectable6 joined #moarvm
10:26 debugsectable6 joined #moarvm
10:35 samcv ok updated nqp tests. kinda weird its coverage for .h files
10:35 samcv https://cry.nu/coverage/
10:35 samcv err ops. my caching is like. too much caching i think? cause i refresh and i don't see a change
10:35 samcv but it is there if i clear cache
10:37 samcv timotimo, https://cry.nu/coverage/coverage/home/samantha/git/MoarVM/src/6model/reprs/MVMNull.h.html
10:37 samcv interesting
10:38 samcv the inlined stuff shows up by file as well as over the entire everything
10:40 timotimo oooh that's kinda cool
10:43 samcv looks like i don't think i have any cache settings in the headers. maybe chrome being weird
10:44 samcv setting expiry at 30 minutes for that folder anyway though
10:47 samcv night all!
10:48 samcv damn you chrome. just stubbornly not reloading the page when i hit refresh :(
10:48 samcv oh well. should be at least better than before. maybe should set cache age lower. since idc about bandwidth
10:49 samcv well it's at 1m now
12:19 lizmat joined #moarvm
15:12 zakharyas joined #moarvm
15:20 zakharyas joined #moarvm
15:42 SourceBaby joined #moarvm
16:08 spebern joined #moarvm
16:11 AlexDaniel joined #moarvm
17:02 jpg51 joined #moarvm
19:16 samcv gonna try out clang/llvm svn just in case llvm-cov has any new awesome changes
19:19 lizmat joined #moarvm
19:37 agentzh joined #moarvm
19:40 timotimo mhm
19:40 timotimo btw: ohai
19:40 lizmat o/
19:40 timotimo i'm writing a new string parser for json_fast
19:44 samcv hello timotimo
19:45 samcv again?
19:45 samcv gonna be every more faster better?
19:45 timotimo hopefully it'll not explode in its memory usage
19:48 timotimo wow.
19:52 timotimo it's "a bit" faster and uses "a bit" less memory
19:52 timotimo but gives equal results
19:53 timotimo huh, but things break a little
19:53 timotimo why is there an escape sequence for / ?
19:57 samcv uhm
19:57 samcv you can escape any character though right?
19:59 timotimo nope
20:00 samcv that's what the rfc says tho
20:00 samcv " Any character may be escaped"
20:02 timotimo huh?
20:02 timotimo json.org says no, but what is that based on again?
20:03 timotimo that's the ecma standard
20:04 timotimo "\h\e\l\l\o"
20:04 timotimo SyntaxError: Unexpected token h in JSON at position 2
20:04 samcv i was looking at https://tools.ietf.org/html/rfc7159
20:04 timotimo that's one parser
20:04 samcv if you escape every character i will hate you
20:04 samcv but. i think it's valid
20:05 timotimo no, i disagree
20:05 samcv kk. well defer to your better knowledge
20:05 timotimo it says any character may be escaped with a six-letter sequence
20:05 timotimo which is the \uAFFE ones
20:05 timotimo "additionally there are a few two-character escape codes"
20:05 timotimo you were giving me a little bit of a shock there
20:05 samcv also asked this in perl6dev but best way to find out which argument to perl6 executable is the filename?
20:05 timotimo anyway
20:05 samcv is it almost always the last one?
20:06 timotimo wanna guess how much better JSON::Fast is now?
20:06 timotimo here's the baseline:
20:06 samcv but it's not always the last one.
20:06 samcv i'm trying to do coverage for roast
20:06 timotimo 12.58user 0.29system 0:13.04elapsed 98%CPU (0avgtext+0avgdata 991276maxresident)k
20:06 timotimo ^- that's with parsing and then rendering
20:07 samcv compared to what?
20:07 timotimo 12.34user 0.27system 0:12.67elapsed 99%CPU (0avgtext+0avgdata 982916maxresident)k
20:07 timotimo without rendering
20:07 timotimo that's the baseline i'm comparing to
20:07 timotimo i.e. what it was on current master
20:07 samcv yay so less. and maybe less ram?
20:07 timotimo um, no
20:08 timotimo it's bad :P
20:08 timotimo but, but, but
20:08 timotimo wanna guess how good it is now?
20:08 samcv uh
20:08 samcv nope. go ahead ;)
20:08 timotimo 0.31user 0.01system 0:00.33elapsed 98%CPU (0avgtext+0avgdata 87228maxresident)k
20:08 samcv i don't believe you
20:08 samcv fully
20:08 timotimo i didn't believe me, too
20:10 timotimo i just pushed it, feel free to play around with it
20:10 timotimo oh crap
20:10 timotimo what did i do wrong
20:11 timotimo huh, zef says it fails a whole bunch of tests
20:11 timotimo but when i run stuff manually it's fine?
20:11 timotimo and prove also says "all tests successful"
20:13 timotimo wtf, zef, what am i doing wrong?
20:34 agentzh joined #moarvm
20:37 samcv try uninstalling it first
20:37 samcv to make sure it's not running the old version
20:37 samcv that's installed
20:38 samcv ahhhh llvm failed a test at the very end of compling everyyyything
20:39 samcv hopefully i can salvage the build
21:03 agentzh joined #moarvm
21:12 timotimo :o
21:28 agentzh joined #moarvm
21:44 agentzh joined #moarvm
21:56 agentzh joined #moarvm
22:04 agentzh joined #moarvm
22:14 agentzh joined #moarvm
22:27 agentzh joined #moarvm
23:08 agentzh joined #moarvm
23:24 lizmat joined #moarvm

| Channels | #moarvm index | Today | | Search | Google Search | Plain-Text | summary