On Fri, Oct 11, 2013 at 8:25 PM, Jeff King <peff@xxxxxxxx> wrote: > On Fri, Oct 11, 2013 at 08:15:46PM -0500, Felipe Contreras wrote: > >> >> You are free to go ahead and implement 'warning ()' in git-sh-setup.sh, in the >> >> meantime no shell script does that, and that's no reason to reject this patch >> >> series. >> > >> > You are completely missing Matthieu's point that we attempt to be >> > consistent in the format of messages, as well as where they are output, >> > and from a user's perspective it does not matter what language the tool >> > is implemented in. >> >> If we truly did that, there should be a warning () function, like in C. > > Or people could hand-code them to look similar, which is exactly what > has happened. And by doing that be prone to make mistakes, like using 'WARNING', instead of 'warning'. But I guess you don't care about consistency _that much_. > If you want to factor out a warning function to clean up the code, be my > guest. But the lack of one does not provide an argument that you should > break consistency. Consistency is already broken. >> > - echo "The configurations pull.rebase and branch.<name>.rebase are deprecated." >> > - echo "Please use pull.mode and branch.<name>.pullmode instead." >> > + echo >&2 "warning: The configurations pull.rebase and branch.<name>.rebase are deprecated." >> > + echo >&2 "Please use pull.mode and branch.<name>.pullmode instead." >> [...] >> >> Are you sure you want me to squash that in? Because the warnings >> wouldn't be consistent. Some would be "WARNING: " and others would be >> "warning: ". Personally I don't care, but if your argument is >> consistency, you should. If we had a warning () function, we could >> truly be consistent. > > It is significant in the most important ways, which are labeling it at > all, and sending it to stderr. Capitalization is less important, in my > opinion. It's still inconsistent. > Using a lowercase version is much more consistent with the warnings > produced by C code, which is why I chose it over the capitalized > version. Again, if you want to change the existing WARNING cases in the > shell scripts to be consistent with C output, be my guest. It seems you are not that interested in consistency after all. > Do you actually have some reason for wanting to output to go to stdout? I'm fine with 'echo "warning: foo" >&2', but still, if you really cared about consistency, there would be a warn() function, precisely to avoid the mistakes of WARNING vs. warning which are already there, plus future ones. -- Felipe Contreras -- To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html