On May 30, 2017, at 8:03PM, Helmut K. C. Tessarek <tessa...@evermeet.cx> wrote:
> On 2017-05-30 17:36, Ryan Schmidt wrote: >> Wouldn't requiring a certain number of reviews make things more >> difficult than they are now, where we don't require reviews? > > Well, this depends. I haven't administered github projects for large > groups, so I don't know if this is even possible with github. > > But I was thinking that new maintainers (with commit access), who still > need to build up a reputation, have to have their PRs reviewed. > Others can commit directly. It is possible to limit who can push to specific branches for a given repo in GitHub. It is possible to require a pull request to push to a given branch in GitHub. It is possible to require positive reviews before a pull request can be merged (or squashed or rebased). > >> It's not allowed to revert a commit that has been pushed. It is >> possible to push a new commit that undoes the changes of a previous >> commit. But MacPorts has additional requirements. For example, if you >> update a port from version 1.0 to version 2.0 and then find that 2.0 >> is broken, you cannot undo it by making another commit that just >> changes the port back to version 1.0 again; you must also increase >> the port's epoch. > > I was rather talking about the process and that it is very easy in git > to rollback commits. > > Let's say commit xyz (v1.1) was bad. > > The maintainer now does the following: > > 1) git revert xyz > 2) makes changes and sets v1.2 > 3) commit > 4) git push > > The last 4 lines were local, so it is ok to have the version changed > temporarily back to pre-1.1 in step 1 (the git revert). Helmut, This is only true if the bad commit xyz was not pushed to the master branch of the macports-ports repo. If it was pushed, then the port's epoch would need to be changed per Ryan's comment. -Sterling > > -- > regards Helmut K. C. Tessarek > lookup http://pool.sks-keyservers.net for KeyID 0xC11F128D > > /* > Thou shalt not follow the NULL pointer for chaos and madness > await thee at its end. > */