IRC channel logs

<gnu_srs1>FYI: N: Missing Signed-By in the sources.list(5) entry for 'http://ftp.ports.debian.org/debian-ports'

<youpi>you can upgrade your apt again to see that much less

<youpi>(2.9.26)

<gnu_srs1>Hi, any idea on how to create /dev/dri/card0 manually for Hurd? MAKEDEV does not have support, and it is not created when starting X.

<gnu_srs1>Due to the xserver-xorg-legacy stuff.

<gnu_srs1>And no device manager.

<damo22>gnu_srs1: you need to write a translator that implements drm

<damo22>i made a start but its complex

<damo22>its basically the graphics driver

<damo22>not just a missing node

<gnu_srs1>damo22: How did you create it?

<damo22>gnu_srs1: i have not completed it

<damo22>its a massive job

<damo22>see the mailing list for discussion on DRM subsystem

<gnu_srs1>damo22: tks.

<gnu_srs1>Still, even if we don't have a translator for drm, the device /dev/dri/card0 should be created by libdrm.

<damo22>not exactly, linux usually exposes that node

<damo22>libdrm is the "user" part of the protocol

<damo22>its already constructed as a server/client model

<damo22>where usually linux provides the server "driver" part and all the api calls go through the ioctl kitchen sink

<damo22>its a compact interface with lots of complexity and thus difficult to reimplement

<damo22>something like 200 ioctls

<Gooberpatrol_66>would it be possible to make an ioctl translator that splits one node into several different files, each for a different ioctl?

<Gooberpatrol_66>i never really liked ioctls

<damo22>err, that would be annoying

<damo22>but the problem actually is that we cannot implement the exact same interface with mig

<damo22>we would need to modify the ioctl syscall i think

<damo22>maybe we need to intercept the ioctl calls and call RPCs instead

<damo22>in libdrm

<damo22>it will be a mess, but probably easier than modifying glibc

<damo22>but the driver needs to come from somewhere

<damo22>maybe we can port it from netbsd

2025-01-30.log