Hi!, Thank you very much for descriptive answer. You are absolutely right. I should read manual more carefully! Still it's quite odd interface. Thank you for your time. Bests, Alex On 10/06/2017 04:32 PM, Jeff King wrote: > On Fri, Oct 06, 2017 at 01:10:17PM +0200, aleksander.baranowski wrote: > >> I'm currently using git version 2.14.2. There is possible to put >> multiple values into same variable with git config. > > Yep, your examples should behave the same even with older versions. > >> Case 1: >> # git config --global user.name Foo - returns 0 >> # git config --global user.name Bar - returns 0 and replace Foo to Bar >> # git config --global user.name Foo - returns 0 and replace Bar to Foo > > This is all as expected. You're hitting the first case in the manpage > synopsis here (I snipped the uninteresting options): > > git config name [value [value_regex]] > > So you're doing: > > git config name value > > which replaces any existing key by default. You could also do: > > git config --add name value > > to add without replacing (if you had a config key that takes multiple > values). > >> Case 2: >> # git config --global user.name Foo - returns 0 >> # git config --global user.name Foo2 Bar - returns 0 and put second name > > Here you're doing: > > git config name value value_regex > > So we're replacing any values which match the regex "Bar" (and there are > none), and leaving others intact. Hence after this you will have two > user.name values. > > If you wanted a name with two strings, you'd have to quote the string > from the shell to leave it as a single argument: > > git config user.name "Foo2 Bar" > >> # cat ~/.gitconfig >> [user] >> email = aleksander.baranowski@xxxxxxxx >> name = Foo >> name = Foo2 > > Right, this is what I'd expect. > >> # git config --global user.name Foo - return 5 and message >> "warning: user.name has multiple values >> error: cannot overwrite multiple values with a single value >> Use a regexp, --add or --replace-all to change user.name." > > And this, too (though I forgot we had such a safety check). > >> It's just an opinion, but this behaviour is no consistent for me. >> >> If it's not the bug it's a feature just let me know. > > It's a feature, though I agree that git-config is rather baroque. We're > mostly stuck with it for reasons of backwards compatibility, though. > > -Peff >
