Esri/developer-support

condemn force pushing

Closed this issue · 9 comments

jgravois commented

we should mention explicity in the readme that pushing your own commits directly to master is strongly discouraged and that force pushing is strictly prohibited, ideally going as far as mimicing the exact rules Level lists for their OPEN open source projects.

i'd even recommend we go a step further

benstoltz commented

I'm 100% on board with that, including your step further.

NoashX commented

Agreed on the first. I think we can better message this internally.

The further step concerns me in case there are ever legitimate reasons to self-publish (e.g. fix README typos). Perhaps we can try again with an email, and then the nuclear option.

Thoughts?

jgravois commented

force pushing is substantially different than a simple push to fix a typo in a file no one else has edited. it is typically only attempted when a normal push fails (either because of a merge conflict or difference in the commit history between the clone and the upstream repository)

git push upstream master
>  ERR
git push upstream master -f
> will succeed and blow away whatever history was present (overwriting with your own)

there is never a good reason for anyone to force push to the master branch of this repository ever and depending on the situation it could be an absolute nightmare to recover from.

the dreaded 'step further' i mentioned is probably not worth discussing, because it appears its only an option for an enterprise instance of github.

nohe427 commented

I'd say just get a few managers with merge access. Other people can still test code and put their thoughts into the issue and when it is ready to be merged in, tag the manager with their @ username call. e.g. @jgravois

jgravois commented

I'd say just get a few managers with merge access

i understand the appeal, but my vote is that we keep it more of an open, open source project.

i think we have enough to gain from an educational perspective to justify the risk.

benstoltz commented

I would agree with @jgravois. However I'd say that if someone abused the open nature of this project that they may have their write access revoked. Obviously one or two mistakes wouldn't justify that response, however if it's repeated and consistent then I think that's an appropriate response. Thoughts?

jgravois commented

i went that step further. because they are 'protected branches', it is no longer possible to force push to either master or gh-pages

phpmaps commented

gone are the days of force pushing. I will sleep better.

NoashX commented

👍