On Sun, Sep 20 2015, David Bremner <david@tethera.net> wrote: > Tomi Ollila <tomi.ollila@iki.fi> writes: > >> make test V=1 (or any other value than 0) and make test V=0 >> works similar way as build in general >> --- >> >> This is 2nd try: In first one I changed test-lib.sh -- there problem >> was the precedence between command line argument and NOTMUCH_TEST_QUIET >> environment variable. Here I don't see a problem there (although >> env NOTMUCH_TEST_QUIET=1 make test V=1 will make environment variable >> overrule the V=1 -- I am not sure who cares, though :D) >> >> Anyway, Someone(TM) may have better solution in mind, so let's >> get bikeshedding going! >> > > One thing I thought of was that we could replace the use of > NOTMUCH_TEST_QUIET with e.g. NOTMUCH_TEST_VERBOSE, and have that set by > the command line argument --verbose. This would be less work on the > Makefile side, but I guess a bit more on the test-lib.sh side. I suppose > it might also be an annoyance for some people currently using > NOTMUCH_TEST_QUIET in some autobuilder setup. At the moment I can't > think of what would really break by changing this "API", but knows. The > current behaviour of having NOTMUCH_TEST_QUIET and --verbose completely > independent is pretty surprising. There was 2 primary motivations when I did that patch (which I forgot to mark rfc although I thought doing so) 1) that was the fastest at the time being and 2) the --verbose is already in use in test-lib.sh. First thing that come to my mind is that if (-v|--v[e[r[b[o[s[e]]]]]]) option is for the suggested purpose its current behaviour is removed and when the option is not give user is informed how to activate "verbose" progress -- but as said that is just first thought... Making NOTMUCH_TEST_QUIET no-op is probably no annoyance to anyone... > > d > > P.S. I hope you appreciate my classic "scope-creep" technique in > bikeshedding. ;) In test-related issues that is fine by me. Tomi