You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
We review shell scripts against standards that live in people's heads. Writing
them down would make review fair and would let us automate the parts a machine
can check.
#789 is the example. The review produced three findings against rules that
appear nowhere in this repo. All three came from one maintainer's private
config. The contributor had followed every rule we actually publish.
Worth deciding and recording:
POSIX sh or bash. Most of bin/ uses #!/bin/sh, but bin/replace and bin/bundler-search use &>, which is a bashism. Pick one and fix the
outliers.
Whether a script should report failure to its caller. bin/clear-port exits
0 when the port was already free, both before and after Clear a port held by more than one process #789. Some scripts
in bin/ use set -e and some do not.
Comment and usage message style.
Whether commit titles follow a keyword list, and if so, which one.
The last point matters most for review. A rule we can point at is a standard.
A rule we cannot is a preference, and holding a contributor to it is unfair.
Related: #793 covers CONTRIBUTING.md, and should link here rather than
repeat any of it.
We review shell scripts against standards that live in people's heads. Writing
them down would make review fair and would let us automate the parts a machine
can check.
#789 is the example. The review produced three findings against rules that
appear nowhere in this repo. All three came from one maintainer's private
config. The contributor had followed every rule we actually publish.
Worth deciding and recording:
shorbash. Most ofbin/uses#!/bin/sh, butbin/replaceandbin/bundler-searchuse&>, which is a bashism. Pick one and fix theoutliers.
bin/clear-portexits0 when the port was already free, both before and after Clear a port held by more than one process #789. Some scripts
in
bin/useset -eand some do not.The last point matters most for review. A rule we can point at is a standard.
A rule we cannot is a preference, and holding a contributor to it is unfair.
Related: #793 covers
CONTRIBUTING.md, and should link here rather thanrepeat any of it.