On Thu, Oct 12, 2017 at 11:10:06AM +0900, Junio C Hamano wrote: > From: Jeff King <peff@xxxxxxxx> > > An earlier patch downgraded "always" that comes via the ui.color > configuration variable to "auto", in order to work around an > unfortunate regression to "git add -i". > > That "fix" however regressed other third-party tools that depend on > "git -c color.ui=always cmd" as the way to defeat any end-user > configuration and force coloured output from git subcommand, even > when the output does not go to a terminal. > > It is a bit ugly to treat "-c color.ui=always" from the command line > differently from a definition that comes from on-disk configuration > files, but it is a moral equivalent of "--color=always" option given > to the subcommand from the command line, i.e. a signal that tells us > that the script writer knows what s/he is doing. So let's take that > route to unbreak this case while defeating a (now declared to be) > misguided color.ui that is set to always in the configuration file. > > NEEDS-SIGN-OFF-BY: Jeff King <peff@xxxxxxxx> Signed-off-by: Jeff King <peff@xxxxxxxx> Thanks for picking this up. I meant to get to it yesterday but ran out of time. Your description looks good to me. > color.c | 17 +++++++++++++++-- > 1 file changed, 15 insertions(+), 2 deletions(-) We should probably protect the command-line behavior with a test. Can you squash this in? diff --git a/t/t6006-rev-list-format.sh b/t/t6006-rev-list-format.sh index 25a9c65dc5..582cab5c8a 100755 --- a/t/t6006-rev-list-format.sh +++ b/t/t6006-rev-list-format.sh @@ -261,6 +261,17 @@ test_expect_success 'rev-list %C(auto,...) respects --color' ' test_cmp expect actual ' +test_expect_success "color.ui=always in config file same as auto" ' + test_config color.ui always && + git log --format=$COLOR -1 >actual && + has_no_color actual +' + +test_expect_success "color.ui=always on command-line is always" ' + git -c color.ui=always log --format=$COLOR -1 >actual && + has_color actual +' + iconv -f utf-8 -t $test_encoding > commit-msg <<EOF Test printing of complex bodies Technically the first test is already covered by the "add -p" we added elsewhere, but I think the sequence make sit easier to understand. Also as an aside, I think this patch means that: git -c color.ui=always add -p is broken (as would a hypothetical "git --default-color=always add -p"). That's sufficiently insane that I'm not sure we should care about it. -Peff
