Re: [kvm-unit-tests PATCH 3/4] build: Make build output pretty

["Nicholas Piggin" <npiggin@gmail.com>]

On Wed Jun 12, 2024 at 8:32 PM AEST, Thomas Huth wrote:
> On 05/06/2024 02.38, Nicholas Piggin wrote:
> > On Mon Jun 3, 2024 at 6:56 PM AEST, Andrew Jones wrote:
> >> On Mon, Jun 03, 2024 at 10:26:50AM GMT, Thomas Huth wrote:
> >>> On 02/06/2024 14.25, Nicholas Piggin wrote:
> >>>> Unless make V=1 is specified, silence make recipe echoing and print
> >>>> an abbreviated line for major build steps.
> >>>>
> >>>> Signed-off-by: Nicholas Piggin <npiggin@gmail.com>
> >>>> ---
> >>>>    Makefile                | 14 ++++++++++++++
> >>>>    arm/Makefile.common     |  7 +++++++
> >>>>    powerpc/Makefile.common | 11 +++++++----
> >>>>    riscv/Makefile          |  5 +++++
> >>>>    s390x/Makefile          | 18 +++++++++++++++++-
> >>>>    scripts/mkstandalone.sh |  2 +-
> >>>>    x86/Makefile.common     |  5 +++++
> >>>>    7 files changed, 56 insertions(+), 6 deletions(-)
> >>>
> >>> The short lines look superfluous in verbose mode, e.g.:
> >>>
> >>>   [OBJCOPY] s390x/memory-verify.bin
> >>> objcopy -O binary  s390x/memory-verify.elf s390x/memory-verify.bin
> >>>
> >>> Could we somehow suppress the echo lines in verbose mode, please?
> >>>
> >>> For example in the SLOF project, it's done like this:
> >>>
> >>> https://gitlab.com/slof/slof/-/blob/master/make.rules?ref_type=heads#L48
> >>>
> >>> By putting the logic into $CC and friends, you also don't have to add
> >>> "@echo" statements all over the place.
> >>
> >> And I presume make will treat the printing and compiling as one unit, so
> >> parallel builds still get the summary above the error messages when
> >> compilation fails. The way this patch is now a parallel build may show
> >> the summary for the last successful build and then error messages for
> >> a build that hasn't output its summary yet, which can be confusing.
> >>
> >> So I agree that something more like SLOF's approach would be better.
> > 
> > Hmm... kbuild type commands is a pretty big patch. I like it though.
> > Thoughts?
>
> Looks pretty complex to me ... do we really need this complexity in the 
> k-u-t? If not, I think I'd rather prefer to go with a more simple approach 
> like the one from SLOF.

The first patch I posted added silent to make too, but I don't love
it because it silences things that you missed or forgot about.

This way is loud by default and you have to adjust recipes to be
quiet. It caught a couple of things I missed the first time around.
I think that's a long term advantage for more short term churn.

Thanks,
Nick