On 1/12/20 4:38 PM, Stephen Rothwell wrote: > Hi Olof, > > On Sun, 12 Jan 2020 11:59:58 -0800 Olof Johansson wrote: >> >> Thanks for the report. Time to automate this at our end, we've had a >> few too many of these slip through all the way to you lately. >> >> Where do you keep your scripts that you catch these things with? Do >> you have a writeup of the checks you do? I should add it to my >> automation once and for all. > > I should export my linux-next scripts as a git repo, but I haven't (yet) :-( > > Attached pleas find check_commits which I run after fetching each tree > and pass the changed commit range. This, in turn, runs check_fixes > (also attached). This is definitively very useful, and after being burned by improper Fixes: tag recently, I ended modifying your check_fixes script and putting in my list of pre-commit scripts to run with a wrapper around, although that required making a few modifications: - add an optional second parameter to let you specify a hook mode which derives the git tree from the current directory - added negative return codes upon error - added a "did you mean" when the SHA1 cannot be found but the subject was correct I am not sure what is the intention of the target_subject vs. subject check as it seems to always fail for me when using a commit that I used as a test for the script, e.g.: subject: net: systemport: Simplify queue mapping logic target_subject: net: systemport: Fixed queue mapping in internal ring map under what circumstances do you have both subjects match? Cheers -- Florian