IRC channel logs

<damo22>demo@zamhurd64:/part3/git/forgejo-runner$ GO111MODULE="on" GOPATH=~/go GO=go-15 GO_FMT=gofmt-15 GOMODCACHE=~/go/pkg/mod make

<damo22>go-15 build -v -tags 'netgo osusergo ' -ldflags '-extldflags "-static" -s -w -X "code.forgejo.org/forgejo/runner/v12/internal/pkg/ver.version=v12.5.3+6-g48edb76c"' -o forgejo-runner

<damo22>internal/pkg/report/mask.go:7:2: package cmp is not in GOROOT (/usr/src/cmp)

<damo22>...

<damo22>how does it resolve "import cmp"

<damo22>do i need to expand all the imports to be absolute ?

<youpi>fyi, I have added a user-agent-based blacklist on darnassus

<youpi>that should help a lot with its load, and thus stability

<nexussfan>Anubis by default allows non-"Mozilla" agents

<nexussfan>and ALWAYS allows git user agentd

<nexussfan>also; I don't think anubis can even run on hurd

<nexussfan>it's go

<youpi>it requires a recent version of go and gccgo doesn't ahve that, yes

<damo22>go-15 is missing cmp, slices and maps builtin modules

<damo22>i tried adding the source

<damo22>iter/iter.go:212:2: use of internal package internal/race not allowed

<damo22>where can i find instructions for cross compiling gnumach?

<damo22>are the only prerequisites having mig?

<youpi>the wiki has a gnumach/building page

<damo22>thanks

<damo22>do you think i should make the CI build mig every commit ?

<youpi>mig doesn't change much, so it's not worth the bother atm

<damo22>debian gnu/linux does not seem to have mig 64 bit in the repo

<damo22>i can build it once i guess and install it

<damo22>it might be too old anyway

<damo22>youpi: i meant the CI for gnumach, as it requires mig, should the gnumach CI build mig first?

<damo22>its not more work to include this if i have to do it anyway to obtain mig

<damo22>i cant easily port the forgejo runner to hurd so i might try to run CI from a gnu/linux system

<nexussfan>Try writing your own runner then

<damo22>how is that easier

<nexussfan>True

<nexussfan>Maybe you could take some things from debian's buildd?

<damo22>it would be a lot easier if the CI was a hurd system but i am having trouble porting the forgejo runner

<damo22>it seems to use more modern go functions than 1.18

<damo22>nexussfan: i think you may be confusing the "runner" with the build script

<damo22>the runner is a client program that connects to a forgejo instance and waits for commands to build the ci

<damo22>then you have a build script that actually tells it what to build during each ci context

<gnu_srs1>sobkas: Congrats for your innovative solution for software rendering of mesa :)

<nexussfan>damo22: Oh

<gnu_srs1>Regarding libdrm your IOCTLs stubs did not work to build it. Can you post these patches again? Thanks.

<damo22>nexussfan: the runner posts back a status of whether the build succeeded or failed and forgejo displays the status next to each commit

<damo22>we may be able to make it mail somewhere the results too

<sobkas>gnu_srs1: I have disabled all of libdrm- intel amdgpu and it was able to compile

<sobkas>Road to having any of them working will make some time...

<nexussfan>damo22: I've used Forgejo actions before

<nexussfan>The runner half-requires docker/lxc anyway

<nexussfan>Which could be replaced with subhurds

<damo22>it doesnt require that

<damo22>its optional i think

<nexussfan>Yes but local running is very discouraged

<damo22>its fine if i control who pushes

<nexussfan>True

<sobkas>gnu_srs1: I built libdrm redirected output to build.log

<sobkas>and done grep '_IOT_.*undeclared' build.log |awk -F' ' '{print "#define " $3 " (0)"}'|sort|uniq

<sobkas>So I had a stubs in no time

<sobkas>and could later do 'real' ones based on this work

<damo22>nexussfan: i can set up a gnu/linux system that runs the forgejo runner and cross compiles mig+gnumach instead and then can run the tests

<nexussfan>That could work

<damo22>so it can have a matrix of tests, for each of i386,x86_64 with and without smp and LDEBUG

<nexussfan>"run the tests" as in running mach/hurd?

<damo22>there is a qemu based test harness

<nexussfan>Okay

<damo22>i havent tried it yet but apparently "make test" might just do it?

<damo22>make check

<nexussfan>I wonder how easy it would be to make a `subdo` command

<youpi>damo22: ? debian does have mig64b

<damo22>ah not on my old version :D

<damo22>i should probably upgrade

<sobkas>mig-x86-64-gnu

<sobkas>So good night

<damo22>im on 11.11

<youpi>that's more than 4yrs old :)

<damo22>yes, because i like avlinux without pipewire

<sobkas>Try experimental/sid

<damo22>early adoption of pipewire was problematic for people who actually record audio

<damo22>its probably fixed now

<sobkas>At least pipewire isn't pulseaudio that one was a lemon

<damo22>yea

<damo22>but firefox still uses the pulse api

<sobkas>That crashes I had, it was wireplumber fault had to downgrade to older version

<sobkas>Api wasn't a problem it was implementation

<damo22>its on my todo list to test pipewire again

IRC channel logs

2026-01-19.log