https://bugzilla.kernel.org/show_bug.cgi?id=212955 --- Comment #14 from Lucas Endres (jaffa225man@xxxxxxxxx) --- (In reply to Alan Stern from comment #13) > THat commit you found, 51e6f07cb12e, cannot be the real culprit. Merge > commits don't change any code (as a general rule). Something must have gone > wrong with the bisection procedure. > > The way to test the result of a bisection is: First, check out the commit > that the bisection identified, build it and boot that kernel, and verify > that the problem occurs. Second, check out the commit's parent (or revert > the commit), build and boot the resulting kernel, and verify that the > problem does not occur. > > One thing you could do is get a list of commits affecting the snd-ua101 > driver and try checking out and testing various ones, to find the first > commit which exhibits the problem. I probably was doing it wrong, then, but thought I was being quite thorough. Here's the procedure I'd been following (probably from https://git-scm.com/docs/git-bisect): 0. cd /usr/src/linux-next 0.1 On the first run, I ran these: "git bisect start", "git bisect bad", "git bisect good $OriginalKnownGoodVersion" (I think the original known good was v5.12.0 or the first commit to 5.12.0) 1. Copy my .config file to the source directory, in case it changes with bisect 2. make -j4 menuconfig, and then just exit and save, just to save .config with any automatic changes needed for this source 3. Remove space hogs and any prior configuration: rm -r /usr/src/linux-next/debian/ /boot/vmlinuz-5.12* /boot/initrd-5.12* /usr/src/linux-image-5.12* 4. make -j4 deb-pkg 5. Install the kernel just built, and update grub accordingly: dpkg -i ../linux-image-5.12*_*.deb 6. Power off the computer 7. Power on and boot the new kernel 8. Test the bug (if its designation changes to "hw:USB1" I expect it'll be working, but still test it): arecord -D hw:UA101 -f S24_3LE -r 48000 -c 2 ./ua101.wav 9. If it records without the error: cd /usr/src/linux-next ; git bisect good, Or with the error: cd /usr/sr/src/linux-next; git bisect bad 10. Repeat 1-10 until I got the output I gave you above. I was under the assumption that running "git bisect {good,bad}" was enough to checkout the next version that needed to be tested, and it seems that it must have been. This is because I had quite a few that didn't work, then one that did and another that didn't, eventually ending with the remaining few all working. If I, indeed, needed to be running checkout (as, say step 0.5), /usr/src/linux-next/.git/refs/bisect contains a logged list of the first bad and some good commits. Reverting each commit previously tested (logged) and retesting sounds even more time consuming, but if you don't mind waiting and think it's necessary, I will continue this endeavor, and eventually get back to you. -- You may reply to this email to add a comment. You are receiving this mail because: You are watching the assignee of the bug.