Linux 4.17-rc3 Released As Another "Pretty Normal" Weekly Test Release

[timofonic]

idk how smart it is to use that one, I just checked out -b amd-staging-drm-next and that tree is based on 4.16.0-rc7. Sure, it's got the latest amdgpu code - but the rest of that kernel ain't exactly up-to-date. If you look back at the logs the mainline kernel hasn't been merged with that tree since 2018-03-25. k10temp can read the temperature on 2400G in 4.17rc3, small things like that would be missing in amd-staging-drm-next along with all the other kernel changes over the last month.

It's a mess. The same happens with the Raspberry Pi Linux kernel, quite outdated compared to upstream. The default branch is based on 4.14.

https://github.com/raspberrypi/linux

I hope new development methodologies, new tools and Git improvements become someday a countermeasure to this kind of forking hell...

[xiando]

You are many of the buggy mainline kernel users, use latests of these: https://cgit.freedesktop.org/~agd5f/linux

idk how smart it is to use that one, I just checked out -b amd-staging-drm-next and that tree is based on 4.16.0-rc7. Sure, it's got the latest amdgpu code - but the rest of that kernel ain't exactly up-to-date. If you look back at the logs the mainline kernel hasn't been merged with that tree since 2018-03-25. k10temp can read the temperature on 2400G in 4.17rc3, small things like that would be missing in amd-staging-drm-next along with all the other kernel changes over the last month.

[davidbepo]

Are these still happening in 4.17-rc3?

https://bugs.freedesktop.org/show_bug.cgi?id=106194

https://bugzilla.kernel.org/show_bug.cgi?id=199505

https://bugzilla.kernel.org/show_bug.cgi?id=199505 is fixed in 4.17rc3 but still happens in 4.16.6

[gukin]

I'm guessing that 2400g is a new critter that just "changes things". I'm finally drinking the kool aid, my 2400G build is being delivered 5/1/2018 and I'll get to experience first hand the frustrations that have dogged the rest of the early adopters of Raven Ridge since February. I was an early adopter of the 7850K chip and had to use the maligned fglrx kernel module until radeon and amdgpu caught up and made it irrelevant; that took several years. I suspect with all the attention paid to AMDGPU that it will be less than "several years" before Raven Ridge is up to snuff; I hope.

[shmerl]

AMD throws random patches to the mainline kernel

Don't they test the result first?

[shmerl]

Mainline kernels do have a partially implemented and buggy amdgpu driver.

How did such buggy result even pass some basic QA? It absolutely breaks normal desktop experience.

[xiando]

Are these still happening in 4.17-rc3?
https://bugs.freedesktop.org/show_bug.cgi?id=106194

I can absolutely positively confirm that freedesktop bug 106194 is totally NOT fixed in 4.17rc3. I get this if the displays go to sleep by power-saving mode, the "kernel BUG at drivers/gpu/drm/amd/amdgpu/../display/amdgpu_dm/amdgpu_dm.c" is still present.

I suspect that this TOTAL SCANDAL is related to why my 2400G can't wake up from suspend correctly in 4.17rc3.

As for https://cgit.freedesktop.org/~agd5f/linux/ - which of those dozen branches would be worth trying?

[shmerl]

Are these still happening in 4.17-rc3?

https://bugs.freedesktop.org/show_bug.cgi?id=106194

https://bugzilla.kernel.org/show_bug.cgi?id=199505

[dungeon]

I wonder what will be the next LTS kernel version.

Likely 4.19 or 5.0 Usually GKH picks whatever comes at the end of the year (of course he also posts his prediction during summer and that is usually last kernel of a year), even PlayStation is best selled... if not everything, during that time

Kind of, from back to school up to the end of year are most interesting times, stronKest quartal+ you know

[xiando]

Linux 5.0 will probably be the LTS version. And 4.17 could be released as 5.0. But then again we could see 4.40 before 5.0.

http://lkml.iu.edu/hypermail/linux/kernel/1804.1/06654.html

we've passed the six
million git objects mark, and that is reason enough to call the next
kernel 5.0. Except I probably won't, because I don't want to be too
predictable. The version numbers are meaningless, which should mean
that they don't even follow silly numerological rules - even if v3.0
and v4.0 happened to be at the 2M and 4M mark respectively.