On Tue, 2019-10-15 at 09:59 +0800, Paul Wise wrote: > Any further thoughts or should I close this bug?
Some further thoughts from me: If you want to change how the run command works, you can do that in your mrconfig, the default just runs the args directly: [DEFAULT] run = "$@" You could just evaluate it as shell code instead, or run a subshell: [DEFAULT] run = eval "$*" [DEFAULT] run = sh -c "$*" Then you would be able to run something like this: $ mr run mr config '$MR_REPO' foo=bar bar=baz You could create a new configure command to mr config the current repo: [DEFAULT] configure = mr config "$MR_REPO" "$@" Then you would be able to run something like this: $ mr configure foo=bar bar=baz myrepos could also change how `mr config` works to make it so that passing only one argument means that `mr config "$MR_REPO" "$1" gets run instead of the current action of returning an error. This would only allow setting one value at a time, make it more convenient to use when already in a repository but could be a surprising footgun when used in a directory tree containing many repositories. $ mr config foo=bar mr config: not enough parameters Another option would be to create a -r / --recurse option that would make normally non-recursive commands like config or register process all repositories instead. The advantage of this would be that it also allows you to configure repos with multiple key=value items instead of just one and the footgun would be hidden behind an explicit option. $ mr -r config foo=bar bar=baz I'm inclined to make these changes to myrepos: * Make the config section optional when the current dir is a registered repos and derive the section from the current dir. * Add a recursive option that would apply to the config and register commands and cause them to do their jobs recursively. In the meantime you should define and use a `mr configure` command. Further thoughts welcome! -- bye, pabs https://wiki.debian.org/PaulWise
signature.asc
Description: This is a digitally signed message part