07:08vedranm: Ristovski: bisect ended on a Linus's merge commit of some media branch with 397 commits
07:08vedranm: anyt tips on how to bisect inside this branch?
07:09vedranm: *any
07:10vedranm: if I understand bisect with merge, it should have already gone throught all of them and this result means that only a combination of commits fails, while them separately do not
07:17ishitatsuyuki: vedranm: bisect with merges is two step. you first bisect down to the merge, then you bisect within the merge
07:19ishitatsuyuki: i don't know how to do the second off-hand, but one way is to bisect on the branch that it was merged from (you will need to find a common base as the good commit)
07:24vedranm: ishitatsuyuki: this? https://stackoverflow.com/a/67951226
07:25ishitatsuyuki: sounds like it
07:28vedranm: ishitatsuyuki: thx
16:28vedranm: ishitatsuyuki: from what I understood, say the merge brings 100 commits, I should manually merge the 50th, and then try it, if it works, I merge 75th, if it doesn't, I merge 25th and on and on?
16:28ishitatsuyuki: no, you bisect on the branch it was merged from
16:28ishitatsuyuki: that's what the SO link you posted do
16:28vedranm: ishitatsuyuki: but that branch works fine, no?
16:28vedranm: hm
16:29vedranm: there's a comment suggesting what I am suggesting now
16:29ishitatsuyuki: let's say some-drm-dev branch is merged into master, you bisect on some-drm-dev
16:29vedranm: since brokencomit^1 and brokencommit^2 both work fine
16:29ishitatsuyuki: huh
16:29vedranm: Yup, both sides work
16:29ishitatsuyuki: that's a merge hiccup then
16:29vedranm: I am testing merge base right now
16:30vedranm: and then I plan to merge half
16:30vedranm: and afterwards either 1/4 or 3/4
16:30vedranm: if that makes sense
16:30vedranm: ishitatsuyuki: I agree
16:30ishitatsuyuki: tell the person who did the merge and they might be able to help you faster
16:30vedranm: that was 2.5 years ago
16:30vedranm: I should test drm-dev on old hardware more often
16:38DottorLeo: Hi!
16:38DottorLeo: Someone here has a RDNA2 card to quickly test a possible bug? ^^"
16:42vedranm: ishitatsuyuki: merge base works, branch brings 109 commits, I am trying merging first 54 or so
16:42ishitatsuyuki: the question is mostly if the merge applies cleanly
16:42ishitatsuyuki: otherwise resolving conflicts by hand is going to be a funny experience
16:43vedranm: that must be
16:43vedranm: if you can merge n commits from the branch, you can merge first k for any k <= n
16:43ishitatsuyuki: that's just not true
16:43ishitatsuyuki: also it's possible that the maintainer resolved conflicts on their own
16:44vedranm: you mean that some mid-state wil not compile?
16:44vedranm: yeah, luckily merge commit is empty
16:44vedranm: so it's a clean merge
16:44ishitatsuyuki: if you have a conflicting commit C in source branch but later reverted, merging will succeed but rebase will not
16:44ishitatsuyuki: you can probably try just cherry picking the range so that you can use automatic bisect
16:44DottorLeo: Fedora 40 KDE , both PC up to date at the same time, one is a desktop pc with a 6750 XT (RDNA2), the other is a laptop with a 5625U (Vega architecture). Using the IKEA planner on the desktop is completely black, on the laptop works great. Mesa is the same version. A possible RDNA bug?
16:45ishitatsuyuki: DottorLeo: probably easier to file a Mesa issue instead
16:45vedranm: ishitatsuyuki: I see what you mean
16:45DottorLeo: yeah, just asking for someone that has another RDNA2 card and test the website
16:45DottorLeo: i'll open a bugreport
16:46ishitatsuyuki: bedtime for me, good luck you both
16:46vedranm:is lucky, there are no reverts
16:46vedranm: ishitatsuyuki: good night, thanks
16:47DottorLeo: night!