|
00:14
cartwright joined
01:27
quotable6 joined
01:28
quotable6 joined
01:30
quotable6 joined
01:48
ilbot3 joined
|
|||
| Geth | MoarVM/coverage: dffbe6593a | (Samantha McVey)++ | tools/update-gh-pages.sh Use coverage repo for coverage reports Should reduce repository size for MoarVM |
04:44 | |
|
05:12
geekosaur joined
|
|||
| Geth | MoarVM/coverage: 672b77e940 | (Samantha McVey)++ | .travis.yml Debug |
05:58 | |
| samcv | at all times. that a lot of the time | 06:01 | |
| sounds neat timotimo | |||
|
06:29
geekosaur joined
|
|||
| Geth | MoarVM/coverage: e29dc1f270 | (Samantha McVey)++ | .travis.yml Fix = that should have been != |
07:08 | |
|
09:51
eater joined
|
|||
| Geth | MoarVM: samcv++ created pull request #581: Merge coverage branch into master |
10:07 | |
| timotimo | right now i get two 404s on moarvm.github.io/coverage/ | 10:09 | |
| i think it needs just a coverage/ in the path before "moar/" and "libmoar/"? | 10:10 | ||
| Geth | MoarVM/master: 18 commits pushed by (Jonathan Worthington)++, (Samantha McVey)++ review: github.com/MoarVM/MoarVM/compare/4...0610b6335d |
10:11 | |
| samcv | yeah. | 10:16 | |
| because it has only build on the 'coverage' branch | 10:17 | ||
| so it built into moarvm.github.io/coverage/coverage/libmoar etc and it has no index.html yet | |||
| i have non master branches build in their own folder | |||
| but don't yet have some automated index file generation for any new branches | |||
| TRAVISSSS | 10:18 | ||
| timotimo | oh, ok | ||
| samcv | builds were super failing | ||
| cause it tried to pull gh/pull/merge+581 or something and said undefined | 10:19 | ||
| wtf travis. it's not that hard to just clone the repository | |||
| and not cause some gibberish | |||
| not clone a non existant thing | |||
| timotimo | urgh | ||
| samcv | so this build may fail. idk i restarted them all | ||
| and now it's starting them all super slowly. so maybe travis is being travis | 10:20 | ||
| timotimo | i think so, yeah | ||
| samcv | ok i think it worked normally this time | 10:21 | |
| also we can test clang 5 if we want to now. i have a new env var for building with the latest clang on the system | 10:22 | ||
| but probably no need cause the coverage branch will build with latest clang anyway. and all the other ones will use the normally installed verision | |||
| also llvm-cov-5 and llvm-profdata-5 were annoying. all these numbered programs | |||
| and then figuring out what exactly they're named on travis is annoying. but it's all good now | 10:23 | ||
| timotimo | i wonder if clang 5 is any faster at compiling interp.c | ||
| samcv | i just have find_latest_version () { ls /usr/bin | grep -Ee "^$1(\-[0-9.]+)?$" | sort -r | head -n 1; } | ||
| Geth | MoarVM: 154e899a9b | (Samantha McVey)++ | tools/update-gh-pages.sh Improve the commit messages of Travis CI to include git describe Also put information on the branch that was built |
11:00 | |
| MoarVM: the-eater++ created pull request #582: Also pass cflags when building probe shared object |
|||
| eater | samcv: ^ | 11:01 | |
| samcv | yay | 11:02 | |
| ok i need to sleep. good night! | 11:06 | ||
| it is way too late | |||
| eater | samcv: sleep well :') | ||
| Geth | MoarVM: cb3f010337 | (Timo Paulssen)++ | tools/update-gh-pages.sh coverage: merge index files into one big index file |
11:37 | |
| timotimo | i hope this works. | 11:38 | |
| eater | that are some nice sed's :'D | 11:39 | |
| timotimo | i don't see a problem with it :P | ||
| eater | I was not saying there was a problem :') | 11:40 | |
| timotimo | it's building your pullrequest before it starts my code | ||
| but wait, how do i trigger it properly ... | 11:41 | ||
| do i have to poke the gh-pages branch? | |||
| maybe we'll want to change the target branch to master? | 11:42 | ||
| eater | timotimo: whats wrong? | 11:45 | |
| I've pushed a commit before you pushed to master | |||
| so first is does my commit | |||
| and then yours | |||
| timotimo | yeah, it's taking ages because travis | 11:46 | |
| eater | idd :( | ||
| timotimo | especially because we have clang | ||
| and clang is like "wow, this file is really big." | |||
| eater | yea | 11:47 | |
| but there is not really a good alternative to travis :/ | |||
| timotimo | is the coverage option an allowed failure? | 11:48 | |
| we should put that in there because otherwise every pull request is going to be rejected | 11:49 | ||
| because it sees it's a pullrequest, prevents the coverage from being made, and that counts as failure | |||
| eater | well its failing on something ele now | ||
| oh | 11:50 | ||
| iv undefined | |||
| if [ "$TRAVIS_PULL_REQUEST" != "false" ]; | |||
| Geth | MoarVM: 86b259fcdd | (Timo Paulssen)++ | .travis.yml allow coverage to fail (because it'll always fail for pull requests) |
11:51 | |
| eater | that is also an option | ||
| timotimo: now lets see | 12:00 | ||
| timotimo | i see it's doing its stuff now | 12:04 | |
| moarvm.github.io/coverage/ | 12:08 | ||
| looks like it works \o/ | |||
| now if there was a little tool that'd indent the html, we could get usable git diff/git log -u output | 12:09 | ||
| eater | timotimo: you can load the html in an XML DOM object and then generate it again with indenting | ||
|
12:12
AlexDaniel joined
|
|||
| timotimo | like, with perl5? | 12:37 | |
| eater | for example | 12:45 | |
| timotimo | feel free to write a script that does that :) | 12:48 | |
| or else i'll just write one that seds a \n after every </tr> | |||
| eater | whoo | 12:51 | |
| my PR now builds succesfully | |||
| timotimo | cool | 13:54 | |
| Geth | MoarVM: cc122eacf4 | eater++ | build/probe.pm Also pass cflags when building probe shared object |
13:55 | |
| MoarVM: 075cd91f92 | eater++ | build/probe.pm pre-define $ENV{CFLAGS} if not defined |
|||
| MoarVM: 1ec6dca832 | niner++ (committed using GitHub Web editor) | build/probe.pm Merge pull request #582 from the-eater/fixNonSharedProbeObject Also pass cflags when building probe shared object |
|||
| eater | thanks nine :D | 14:02 | |
| dogbert17 | timotimo: you around? | 15:08 | |
| timotimo | yup | ||
| dogbert17 | I have a stupid c question :) | ||
| timotimo | i know all about stupid c | ||
| dogbert17 | isn't the test at line 307 unnessecary? github.com/MoarVM/MoarVM/blob/mast...ray.c#L301 | 15:09 | |
| timotimo | yeah, uint and < 0 | ||
| it's a bit unfortunate that we are mixing uint and int all over the place | 15:10 | ||
| but moar currently doesn't really have uint at the bytecode level | |||
| geekosaur | turn on blame view nd chase the commits and you'll likely find it used to be int | ||
| dogbert17 | I'm cppchecking again btw but my c fu is not up to snuff hence the questions :) | 15:11 | |
| nine wonders whether it'd be possible to cache lookups into MVMP6opaqueNameMap. Considering that each attribute access seems to loop through this map to find the attribute slot. | 15:12 | ||
| But probably spesh and or the JIT already take care of that | |||
| dogbert17 | here's another thing I can't wrap my head around: look at the pieces of code starting at 1333 and 1354, are the the same ? github.com/MoarVM/MoarVM/blob/mast...ph.c#L1333 | 15:16 | |
| eater | dogbert17: attr vs attrs? | 15:18 | |
| dogbert17 | eater++, didn't see that :) | 15:20 | |
| eater | :p had to look for a minute myself too | 15:21 | |
| timotimo | i now have a perl6 script that grabs all the coverage percentages from the html files for every commit in coverage/gh-pages | 15:36 | |
| eater | nice :) | 15:38 | |
| timotimo | it's also rather fast | 15:50 | |
| hm, so when i graph stuff, do i graph percentage over time, or just pieces covered over time | 16:48 | ||
| i.e. when we stay at 100% coverage, but more stuff is covered, should the line stay or go up? | |||
|
17:03
TimToady joined
|
|||
| samcv | wow nice timotimo moarvm.github.io/coverage/ fancy | 19:40 | |
| timotimo | oh! i was just wondering when you'd show up today :) | ||
| samcv | very good :) | ||
| timotimo | i made something | ||
| samcv | all on one page is very nice | ||
| timotimo | gist.github.com/timo/437c24c85c8c5...a681f00538 | 19:41 | |
| point it at the .git of your local clone of moarvm/coverage for the "extract" command, then run it with "graph", and it'll create a crapton of png files in your CWD | 19:42 | ||
| oh, also, it requires an "analysis" folder in its CWD | |||
| samcv | wow | 19:43 | |
| timotimo | i was hoping we could set these graphs as background images in the table | ||
| samcv | also make sure to not include anything in folders not named `moar` or `libmoar` cause those are probably gonna be other branches that had covenage built | 19:46 | |
| timotimo | it doesn't look into folders at all | 19:47 | |
| it reads the index files | |||
| samcv | ah ok | ||
| yeah | |||
| timotimo | at the moment the data is very boring, though | 19:48 | |
| samcv | yeah i'm sure | ||
| timotimo | there's only 4 dates that have data at all | ||
| and only 3 fields in the whole grid have any change in them | |||
| strings/utf8_c8.c-line => (31.65 30.96 30.96 30.96) | |||
| 6model/reprs/VMArray.h-region => (82.73 82.42 82.42 82.73) | |||
| strings/utf8_c8.c-region => (22.66 21.88 21.88 21.88) | |||
| samcv | also what we could do at some point is to do roast coverage and add it to nqp coverage | 19:49 | |
| timotimo | that'd be neat | ||
| samcv | if we want to make this test an ignored failure and use fast_finish i *think* though i'm only guessing what that option does for rakudo, that it doesn't wait for those allowed failures to finish before saying it's success or failure | ||
| or could go insane and do it per roast test file | 19:50 | ||
| since it creates files in ./coverage/t/01-blah/blah.t.profraw | |||
| etc for all the test files | |||
| timotimo | we could do it when somebody asks for that particular data | ||
| samcv | yeah | 19:51 | |
| timotimo | but i'm not sure how we're supposed to present the data we could make from that | ||
| samcv | idk a link to all the coverage for all the test.t files? | ||
| would have to be a link | |||
| timotimo | i could imagine we could make it in three stages | ||
| samcv | to ./overkill directory XD | ||
| timotimo | at the outermost we get the data for nqp testsuite + rakudo testsuite | 19:52 | |
| samcv | yah | ||
| timotimo | then you can get data for each folder in roast as well as each folder in t/ | ||
| and then you can see data for each individual file | |||
| samcv | though combined nqp testsuite + rakudo testsuite is gonna be more useful than the t/blah files for now. | ||
| probably want to save that for a later date | |||
| because it will make a lot of fucking files | 19:53 | ||
| timotimo | mhm | ||
| samcv | oh timotimo i can output data as json | ||
| for coverage. but i don't know wtf it outputs. but let me check | |||
| timotimo | well, too late, i already parse all the html | 19:54 | |
| samcv | heh | 20:01 | |
| well. i'm just curious what data it gives | |||
| if i can make this program work >:| | |||
| i added you ability to push to github.com/samcv/MoarVM-cover/ though. which is where the things that generate the coverage are kept | 20:03 | ||
| timotimo | cool | 20:05 | |
| thing is: i've built this in perl6 | |||
| samcv | that's ok | ||
| we can build perl6 for the coverage branch | |||
| or... use an appimage | |||
| timotimo | bleh. i'd rather we grab an appimage or so | ||
| samcv | XD | ||
| timotimo | yeah | ||
| samcv | yeah | 20:06 | |
| 10MB dl and none other things to setup | |||
| timotimo | yeah | ||
| and it should immediately work with JSON::Fast, too, right? | |||
| samcv | just know you need to cd into %*ENV<OWD> at the start | ||
| yepu | |||
| timotimo | except i'd really like the newest JSON::Fast ... | ||
| samcv | just cd into OWD | ||
| hm | |||
| timotimo | or just throw out the intermediate step | ||
| it doesn't have to roundtrip to disk if we can't keep old data around anyway | 20:07 | ||
| samcv | it will still work with older JSON::Fast right? | ||
| timotimo | gnuplot is driving me just slightly mad | ||
| yeah, it'll potentially be slowre, though? | |||
| samcv | yeah. nobody will be injured by slowness though | ||
| timotimo | gah, gnuplot | 20:10 | |
| isn't cooperating | |||
| you know gnuplot? | 20:11 | ||
| samcv | i know of it. but have hardly ever tried to use it | ||
| timotimo | oh, that's how it works | 20:12 | |
|
20:14
committable6 joined
|
|||
| samcv | also DATE_VERSION_HEADER="$(printf "%s\n%s" "$(date --utc)" "$(./nqp --version 2> /dev/null)" )" | 20:15 | |
| this is what i use to generate the date_version_header (though env var won't be availible outside the script) | |||
| but you can do DATE_VERSION_HEADER="$(printf "%s\n%s" "$(date --utc)" "$($NQP_FOLDER/nqp --version 2> /dev/null)" )" | |||
| timotimo | ok, i got a filled curve now | ||
| samcv | since we now have a $NQP_FOLDER variable in travis.ci | ||
| to put at the top of the file | 20:16 | ||
| timotimo | what should the curves look like? | ||
| samcv | .uh. | ||
| timotimo | i want to put it as the background for each cell in the overview grid | ||
| samcv | uhm | ||
| dunno. what do you have now | |||
| what are the choices? | |||
| timotimo | we can choose the line color, color above and below, transparent is an option, and patterns are also available, as well as half-transparency | 20:17 | |
| what kind of screen resolution should we be targeting? | 20:19 | ||
| samcv | uhm | ||
| timotimo | because when we have these pngs, we need to give each cell a fixed size | ||
| samcv | :O | ||
| they're already all the same size | 20:20 | ||
| well sorta | |||
| timotimo | "sorta" isn't gonna cut it :) | ||
| samcv | idk i haven't measured | ||
| sec | |||
| only the left most one is a diff size | 20:22 | ||
| about 165 px instead of 182 | |||
| can't we just have them stretched to fit the size | |||
| but i guess we can change the size or whatever | 20:23 | ||
| also we have a css file | |||
| moarvm.github.io/coverage/style.css | |||
| timotimo | it'd look terrible :D | ||
| samcv | why? | ||
| make them svg? | |||
| timotimo | hm, it's possible | 20:24 | |
| the svg files are like 10x as big | 20:25 | ||
| in bytes | |||
| samcv | hm | ||
| timotimo | and i'm expecting to put all the png files into one big file and optipng the heck out of it | 20:26 | |
| samcv | well. can we start with svg files and not messing with the size of the cells | ||
| and then see how they look | |||
| and can then decide how to size the cells | |||
| or would that be to omuch work | |||
| timotimo | it's at least one bunch of work | 20:27 | |
| no, two bunches actually | |||
| another option is to have a small image on the right of each cell | |||
| now how should that look so it works with all background colors ... alternatively, i can know what background color each cell will get | 20:30 | ||
| in that case i might want just a solid line instead of filled curve | 20:32 | ||
| yeah, that'll be fine. | 20:33 | ||
| i've montaged all the lines into one png file and optipng'd it, now it's only 867 bytes big | 20:35 | ||
| that can go into a .css file as a data: url | |||
| timotimo tries some trickery | 20:45 | ||
| samcv | well the colors depend on the coverage | 20:54 | |
| timotimo | of course | 20:56 | |
| i know the numbers, though | |||
| so i know the color | |||
| now to put img tags everywhere ... | 20:58 | ||
| how annoying. | 21:04 | ||
| vim macro to the rescue, hopefully | |||
| samcv | hah | 21:08 | |
| timotimo | nah, i search-replaced it, but how do i do the thing with the right numbers :\ | 21:09 | |
| samcv | do the thing? | 21:12 | |
| oh | |||
| just check their shading | |||
| ? idk | |||
| you wanted the numbers on top of the graphs right? | |||
| timotimo | not necessarily | ||
| did you see irclog.perlgeek.de, the main page? | 21:13 | ||
| i'm having a hard time making it obvious where the 0 line is | 21:20 | ||
| the heck did i do now? o_O | 21:22 | ||
| i was getting a bit tripped up by the "- (0/0)" entries | 21:25 | ||
| this ain't half bad | 21:29 | ||
| except i'm still tripping over the "- (0/0)" entries for some reason :\ | 21:31 | ||
| aha! | 21:32 | ||
| there's no space in front at all | |||
| hm, do i get one for TOTALS at all ... | 21:33 | ||
| nope, it doesn't have an <a> tag so i'm skipping it | 21:34 | ||
| i shouldn't | |||
| imgur.com/a/DKzJj - samcv check it out! | 21:36 | ||
| i should fake up some values there %) | 21:38 | ||
| yup, fake random data coming right up | 21:40 | ||
| same link | 21:41 | ||
| added one more file with 60px wide graphs instead of 30px, also with a space between image and text | 21:45 | ||
| gist.github.com/timo/437c24c85c8c5...a681f00538 - my code again | 21:48 | ||
| with the legit data, where it hardly changes at all, the image gets optimized down to 2 kb, with noise data it goes up to 64kb, which is still 20kb less than the html file itself | 21:55 | ||
| samcv | gonna get some tea | 22:09 | |
| oh mah gawd | |||
| beautiful | |||
| timotimo | isn't it! | 22:11 | |
| i'm just now modifying it so that it can also accept html files where the img tags are already put in | |||
| samcv | gotta remove the Generated by llvm-cov and replace with Generated by llvm-cov and *magic* | ||
| timotimo | oooh, is there an emoji that's basically magic sparkles? | 22:12 | |
| might continue working on this once we have some actual data points :) | |||
| like when you add rakudo's test folder or maybe even roast to the mix | 22:13 | ||
|
22:48
AlexDaniel joined
|
|||
| timotimo | samcv: i wish i had the same kind of energy for writing that C code in UCD %) | 23:57 | |
| but for now i'll sleep soon | |||
| samcv | me too | ||
| kk :) | |||
| thank you for magic! | 23:58 | ||
| timotimo | it'll be much more fun to write that code when you get paid :) | ||
| community support seems strong for your application | |||
| samcv | very much so | 23:59 | |
| hahaha | |||