=== cascardo_ is now known as cascardo
EickmeyerHi Kernel Team! I had a bit of a suspicion that the lowlatency kernel was split into a different source and bug 1999528 confirmed it. As such, and that SRU as exhibit, I'm suspicious that it's no longer getting the same amount of attention and quality it once had. 14:44
-ubottu:#ubuntu-kernel- Bug 1999528 in linux (Ubuntu Kinetic) "[DEP-8] Run ADT regression suite for lowlatency kernels Jammy and later" [Medium, In Progress] https://launchpad.net/bugs/199952814:44
EickmeyerThere have also been a number of user complaints about the quality in mailing lists and reddit posts, so I'm getting a bit worried here.14:45
EickmeyerOne such thread: https://lists.ubuntu.com/archives/ubuntu-studio-users/2022-December/011816.html14:46
EickmeyerSo, if anybody can ease my mind, that would be wonderful. :)14:46
mhcerriEickmeyer, it was split just to make the build process easier for us. but that should be transparent for users. binary packages are still the same afaik, and we are handling it as a derivative kernel which means it receives the same fixes as the generic kernel and the same level of testing15:06
mhcerriklebers, xnox, anything to add about that? ^15:06
Eickmeyermhcerri: Thanks for that. In Kinetic, at least, the version numbers are different, so that's where I first noticed the split. But where it's really concerning is that users are noticing regressions when it comes to low latency when working with audio devices as they have become accustomed to. I.e. pressing a key on a midi keyboard and havine15:20
Eickmeyerinstantaneous response in their speakers.15:20
EickmeyerThose issues might be upstream, but I find it coincidental that they happened around the same time as the split.15:21
EickmeyerFWIW, I'm usually the last person to blame the kernel. :)15:23
mhcerriEickmeyer, I agree. it's worth it checking for sure. let's see what the others (above) say15:24
=== cpaelzer_ is now known as cpaelzer
klebersEickmeyer, as mhcerri mentioned, we have split the lowlatency flavor to a separate derivative due to build times getting extremely large on some architectures as the flavors are built in series. As for the quality, it should be the same as before, this kernel will always get the latest updates from the generic kernel and we run the same regression tests as before. Actually, being a separate derivative makes it possible to run ADT on it as well, 16:06
klebersas for ADT only the generic flavor is tested. 16:06
klebersso it should be just a coincidence that users are finding issues now16:06
Eickmeyerklebers: OK, that puts my mind at ease with the split, but does make me wonder about regressions wrt the entire lowlatency config stack and if that needs to be evaluated.16:09
klebersEickmeyer, being a separate source really changes how we manage the configs, at least for now. What kind of config stack regression are you worried about?16:25
Eickmeyerklebers: I'm not sure, as I don't consider myself a kernel expert, however perhaps some reevaluation of the PREEMPT flags needs to be considered. The "crackling" that users are mentioning are almost certainly xruns in the buffer where they didn't exist before, and no parameters were changed on their part.17:09
=== sforshee_ is now known as sforshee
=== tomreyn_ is now known as tomreyn
=== cengizIO is now known as cengiz_io

Generated by irclog2html.py 2.7 by Marius Gedminas - find it at mg.pov.lt!