Hi, On Fri, Jun 21, 2024 at 6:01 AM Michael Gofron <gofronm@xxxxxxxxx> wrote: > > Hello, > > My question is can using `git bisect skip` cause a bisect to be indeterminate and/or fail if the commits that are skipped couldn't have caused the issue? If some skipped commits haven't caused the issue but they are direct ancestors of the commit that caused the issue, then git bisect might not be able to tell which one among those commits and the commit that caused the issue is the actual commit that caused the issue. For example if there is the following history: G1-S1-G2-G3-S2-S3-B1-S4-S5-B2-B3-S6 (where G* are "good" commits, S* are skipped commits and B* are "bad" commits) then git bisect will not be able to determine which one is the first bad commit between S2, S3 and B1. > Consider if my commits are like this: > > 1P - 2P - 3B - 4P - 5P - 6B - 7B - 8F - 9F. > P for "Pass", B for "Broken", and F for "Fail". > Broken commits are commits that we can't create a build for but wouldn't cause the issue. > Failing commits are failing because of a bug. Similarly as the case above, I would say that git bisect will not be able to determine which one is the first bad between 6B, 7B and 8F. > In this case, 8F caused the bug. > If you tell git bisect that 1P is good and 9F is bad, bisect picks a commit between the known newest Good commit (1P) and the known oldest Bad commit (9F). Yeah, it should pick one in the middle, so likely 4P, 5P or 6B. > 1P -- 2P - 3B - 4P - 5P - 6B - 7B - 8F - 9F > G B > <------------------------------> > Perhaps 4P. That builds and passes, so it marks that as Good. > > If it then goes to 6B which is a Broken commit and we do `git bisect skip` what happens next? It seems from the code it uses a psuedo random number generator with bias to determine the next commit. Yeah, because it tries to avoid testing broken commits that are likely to be around the broken commit it already picked, while at the same time it tries to be efficient which means to pick a commit near the middle of the range of commits left. > Would it ever get in a state where it can't determine the commit that caused the issue even if these broken commits would never cause an issue? It can't know that the broken commits that are direct ancestors of a "bad" (or "failed" commit as you call them) would never cause the issue, so it should eventually say that it can't determine which commit is the first bad commit among 6B, 7B and 8F. If you are sure that some broken commits don't have the issue you can mark them as "good" to help git bisect which should then be able to find the first "bad" commit. Best, Christian.
