timo the "run full spec tests with spesh nodelay" takes almost exactly one hour 00:10
ugexe i used to have a pipeline that ran spectest using prove. dunno if it would still work, or if you actually want to use prove 00:18
github.com/ugexe/perl6-build-pipel...le#L72-L88
timo can you tell me what the difference is between TAP::Formatter::JUnit and JUnitREGRU? 00:24
there is apparently also TAP::Harness::JUnit 00:26
ugexe if i remember correctly it was that something in jenkins (or maybe the variant of jenkins i was using, hudson) didn't support junit correctly 00:28
::Formatter and ::Harness are for different things
Geth rakudo/azure_improvements: 6dda49b4e7 | (Timo Paulssen)++ | azure-pipelines.yml
For now no builds, just test some perl5 stuff
00:35
ugexe github.com/regru/p5-TAP-Formatter-...7f226b0c78
looks like the main difference in related to encoding
timo > The only difference between this module and TAP::Harness is that this module adds the optional arguments 'xmlfile', 'package', and 'namemangle' that cause the output to be formatted into XML in a format similar to the one that is produced by the JUnit testing framework.
that's from TAP::Harness::JUnit 00:36
TAP::Formatter::JUnit* is usable when you need JUnit output from a different TAP::Harness implementation, i imagine?
Geth rakudo/azure_improvements: 3995f3b95e | (Timo Paulssen)++ | azure-pipelines.yml
fudge out extra moar stability tests for now
00:39
ugexe yeah im not sure. maybe it predated the plugin framework, or the author wasn't aware of the plugin framework? metacpan.org/pod/TAP::Harness#WRITING-PLUGINS seems to suggest writing a formatter makes sense 00:40
timo right 00:42
you know perl5 too right? i don't, so if you could hand-hold me for a brief moment i would appreciate it
ugexe yeah
timo cool. ok so i want to install TAP::Formatter::JUnit, but on the linux image when i "cpan install" that i get "if you answer no to the question below we set up everything for you automatically", is it safe to just `echo no | cpan install blabla`? on the windows runner instead i get "Can't locate CPAN/Author.pm in @INC (you may need to install the CPAN::Author module) [...]" but i think that's 00:44
because i used "bash" instead of "pwsh" where i would get strawberry perl instead
ugexe use cpanm instead 00:45
Geth rakudo/azure_improvements: b0f0232f5a | (Timo Paulssen)++ | azure-pipelines.yml
run pwsh on windows instead of bash
rakudo/azure_improvements: df7a2722d1 | (Timo Paulssen)++ | azure-pipelines.yml
tell cpan no.
timo do i have to install that first with the "wget cpanmin.us + perl that_file" method?
ugexe ive used perl for a long time and i dont think i ever bothered figuring out 'cpan'
i think some perl builds might just come with cpanm now? for example strawberry perl. but if not then yeah 00:47
timo i'll give that a shot
do you think there's a simple way to get the harness to also output special azure-compatible percentage-completion lines to stdout or stderr every minute or two or even every x lines? 00:48
Geth rakudo/azure_improvements: b08b1f1e64 | (Timo Paulssen)++ | azure-pipelines.yml
use cpanm instead of cpan (see if it exists)
00:50
ugexe i'll see if i can find something, but my immediate search failed. i thought there was some ::Tee type plugin that allowed me to output tap to stdout while also saving JUnit to files that the CI system understood 00:51
timo Successfully installed TAP-Formatter-JUnit-0.16 00:53
6 distributions installed
there's an env var that outputs the raw tap stream into a file as well
but that might only happen after the fact 00:54
worst case, just piping the output from the reporter through a simple perl script that counts lines against the number of non-commented lines in the spectest.data file that's in use could be good enough 00:56
ugexe yeah ::Formatter::JUnit 01:06
btw with cpanm you don't use 'install'. so just `cpanm TAP::Formatter::JUnit` i believe 01:13
timo oh ok thank you 01:17
ugexe TAP::Formatter::JUnit also comes with a tap2junit command that can turn TAP into JUnit after the fact 01:26
timo oh, that could be useful i think for the "make m-spectest" command that isn't set up to accept and pass on an argument for what formatter to create 01:33
ugexe i would guess prove has environment variables for setting options 01:34
"Default options may be set by specifying the PROVE_SWITCHES environment variable."
timo it's a custom harness script, not using prove 01:36
ugexe ah 01:40
timo hm. so on the mac runner it's installed App::Cpanminus successfully, but doesn't tell me where, so "cpanm" after the installation results in "command not found" 01:50
! Installing the dependencies failed: Module 'Test::DiagINC' is not installed 01:55
huh, what happened there
"See /root/.cpanm/work/1726451316.1966/build.log for details" well that's long gone ... 01:56
ugexe i think if you run with -v it will output whatever would be in build.log 04:18
07:29 Geth joined 07:30 lizmat_ left, lizmat joined
Geth rakudo/azure_improvements: f758051371 | (Timo Paulssen)++ | azure-pipelines.yml
ls recursively the test results folder
07:32
08:00 sena_kun joined
Geth rakudo/azure_improvements: cfbc7f27af | (Timo Paulssen)++ | azure-pipelines.yml
look in correct folder for test results ...
08:28
timo test success rate now shows up on the azure pipeline dashboard 08:51
they are not grouped, however 08:54
Geth rakudo/azure_improvements: 4ecc441df4 | (Timo Paulssen)++ | azure-pipelines.yml
run smaller spec test and upload all resulting files
09:09
rakudo/azure_improvements: 149fb979a2 | (Timo Paulssen)++ | azure-pipelines.yml
give test run a distinct name per instance
09:18
rakudo/azure_improvements: 3c0f8baee7 | (Timo Paulssen)++ | azure-pipelines.yml
also publish test results folder for regular tests
09:29 sena_kun left
timo there's a moarvm panic in this test run on mac and the harness doesn't seem to output anything that looks like a failure, but the exit code is nonzero at least. odd. 09:39
09:45 librasteve_ joined
Geth rakudo/azure_improvements: dc8c81ac22 | (Timo Paulssen)++ | azure-pipelines.yml
fix smallification of spectest.data
10:06
rakudo/azure_improvements: 8a3f043431 | (Timo Paulssen)++ | azure-pipelines.yml
spectest.data lives in a subfolder oops
10:18
rakudo/azure_improvements: d3233e1aec | (Timo Paulssen)++ | azure-pipelines.yml
try to fix tap2junit invocation
10:29
tbrowder hi, where can i find specs on new pod/rakudoc? 10:59
finanalyst: ^^ 11:00
tellable6 tbrowder, I'll pass your message to finanalyst
tbrowder ah, got it: github repo: Raku/RakuDoc-GAMMA 11:08
|Tux| Rakudo v2024.08-60-gdde756878 (v6.d) on MoarVM 2024.08-9-g570db2612
csv-ip5xs0.262 - 0.265
csv-ip5xs-201.090 - 1.100
csv-parser1.508 - 1.511
csv-test-xs-200.141 - 0.141
test1.792 - 1.882
test-t0.401 - 0.403
test-t --race0.256 - 0.260
test-t-204.808 - 4.828
test-t-20 --race1.154 - 1.175
11:09
tux.nl/Talks/CSV6/speed4-20.html / tux.nl/Talks/CSV6/speed4.html tux.nl/Talks/CSV6/speed.log
Geth rakudo/azure_improvements: 6d60d5196d | (Timo Paulssen)++ | azure-pipelines.yml
use find instead of wonky asterisks
11:40
ab5tract those wonky asterisks are awesome in fish shell :) 11:51
but also ./foo/**/* would be redundant in fish 11:52
timo does bash have a **/*(but-only-plainfiles)? 11:54
i know zsh does have something like that
ab5tract too many globs are bad for your teeth, or so I've heard 11:55
but I assume that your shift to find over the double asterisk glob indicates that not all was working well before 12:04
Geth rakudo/azure_improvements: e5831abf1e | (Timo Paulssen)++ | azure-pipelines.yml
xml editing with sed? what could go wrong!
12:57
rakudo/azure_improvements: 0b40896a19 | (Timo Paulssen)++ | azure-pipelines.yml
fix copypasto, also apply to regular test jobs
13:00
rakudo/azure_improvements: f276a19a26 | (Timo Paulssen)++ | azure-pipelines.yml
strip initial ./ to prevent class names starting with ..
13:05
rakudo/azure_improvements: bd83ee9999 | (Timo Paulssen)++ | azure-pipelines.yml
fix working for test file edit script
13:13
rakudo/azure_improvements: 290b196358 | (Timo Paulssen)++ | azure-pipelines.yml
strip .junit.xml from paths as well
13:17
timo ok i'm not sure why azure isn't showing the test as groups or a tree or something like that 13:23
Geth rakudo/azure_improvements: 62e01edc93 | (Timo Paulssen)++ | azure-pipelines.yml
need to run this code as bash not script (for windows)
13:34
timo the "group by" dropdown seems to appear once the run is finished, or something silly like that? 13:36
Geth rakudo/azure_improvements: e2bb0449a9 | (Timo Paulssen)++ | azure-pipelines.yml
try --merge in prove, see if that gets us moar panics in our files
13:47
timo there's an moar panic that i can't attribute to the test file it occured in :| 13:48
it kind of looks like --timer on prove with --formatter=TAP::Formatter::JUnit gives me the time it takes to parse the TAP output and create the junit xml which is in total very fast 14:07
Geth rakudo/azure_improvements: c769882ade | (Timo Paulssen)++ | azure-pipelines.yml
try to install and use rr, to see if it would even work
14:26
rakudo/azure_improvements: 6d90ab07cc | (Timo Paulssen)++ | azure-pipelines.yml
you gotta sudo apt
14:56
15:35 librasteve_ left
timo is 06-telemetry/02-usage.t known to sometimes crash? 15:56
it's possible it's doing something that's not threadsafe
it's happening on our mac runner, so maybe i shouldn't be surprised that i can't reproduce it locally 16:12
Geth rakudo/azure_improvements: 849a5910c1 | (Timo Paulssen)++ | azure-pipelines.yml
try rr record -n and setting perf_event_paranoid. probably not allowed tho
16:16
rakudo/azure_improvements: cd39226298 | (Timo Paulssen)++ | azure-pipelines.yml
try latest stable ubuntu for rr with more hardware support
16:28
ab5tract timo: not 100% sure but I do semi-recall that being a bit of a flapper.. 16:32
is the mac runner on intel or aarch64?
timo: it passes fine locally with latest HEAD (`xnu-10063.121.3~5/RELEASE_ARM64_T6020 arm64`) 16:37
timo ab5tract: there are arm64 images available, but not for the macOS-12 image that we're using in this pipeline 16:49
for VM related reasons i can't "rr record" on the runner, sadly. it would have been dope to just have a little influx of replayable crash recordings to look at whenever we want 16:50
but i guess i'll have to settle for coredumps of panics and oopses and such 16:52
ab5tract I don’t see any obvious reason to stick to an old macOS image. Am I missing something? 16:59
Geth rakudo/azure_improvements: 14b9e84bc9 | (Timo Paulssen)++ | azure-pipelines.yml
see if coredumpctl is set up for us to use
17:05
timo *shrug* i guess apple is pressuring people to update their OS more than microsoft does? 17:07
ab5tract I've never had macOS just apply an update the way that Windows will just suddenly restart in the middle of whatever I'm doing :) 17:15
but 2 major versions is a while back 17:16
Geth rakudo/azure_improvements: b6125e7f9a | (Timo Paulssen)++ | azure-pipelines.yml
try getting crashes out of apport
17:22
rakudo/azure_improvements: b02ecd16fb | (Timo Paulssen)++ | azure-pipelines.yml
paths path paths
17:25
timo cool, core dumps. 17:30
Geth rakudo/azure_improvements: 78771eda3e | (Timo Paulssen)++ | azure-pipelines.yml
make success of prev stage a condition for later stages
17:34
rakudo/azure_improvements: c5123fae4a | (Timo Paulssen)++ | azure-pipelines.yml
check out apport-retrace
rakudo/azure_improvements: 6bcd7058b6 | (Timo Paulssen)++ | azure-pipelines.yml
seems like i have to install apport-retrace first
17:36
rakudo/azure_improvements: a8bbae8e94 | (Timo Paulssen)++ | azure-pipelines.yml
oops wrong variable
17:43
rakudo/azure_improvements: 7be9cf78c9 | (Timo Paulssen)++ | azure-pipelines.yml
rebuild that package info, apport-retrace, please
17:56
18:10 sena_kun joined 19:30 vrurg joined 19:33 bloatable6__ joined 19:34 tellable6__ joined, greppable6__ joined 19:35 andinus2 joined 19:40 bloatable6 left, andinus left, greppable6 left, vrurg_ left, tellable6 left 20:34 [Coke]__ left 20:53 [Coke]_ joined
Geth rakudo/azure_improvements: 31ac6d9cad | (Timo Paulssen)++ | azure-pipelines.yml
check apt sources, and also look if apport is set up on 24.04
20:55
21:06 [Coke]_ left 21:19 [Coke]_ joined 21:29 [Coke]_ left 21:40 sena_kun left 22:01 [Coke]_ joined