On Donnerstag, 8. Oktober 2020 15:21:54 CEST Paolo Bonzini wrote: > On 08/10/20 15:09, Christian Schoenebeck wrote: > >> But doesn't this (and patch 6 as well) break TAP output? Using > >> g_test_message + g_test_verbose would be the best of both worlds. > > > > If there was TAP output then yes, patches 4, 5, 6 would probably break it. > > > > How/when is TAP output enabled? I don't see any TAP output by default. > > With "--tap", but with glib 2.62 it will be enabled by default. For > example on Fedora 32: > > $ ./test-mul64 > # random seed: R02S3efb20d48a41e1897cb761e02393c11b > 1..2 > # Start of host-utils tests > ok 1 /host-utils/mulu64 > ok 2 /host-utils/muls64 > # End of host-utils tests > > I'm okay I guess with using g_test_message on 2.62 or newer, and > assuming people don't use --tap --verbose on older versions.
Simpler solution: just appending '#' character in front of each printf() line, that would be both fine for TAP and regular output: http://testanything.org/tap-specification.html#diagnostics Unfortunately 'test_tap_log' is a private variable in glib (gtestutils.c), otherwise I could have made that conditionally. There is no getter function in the glib API for this (TAP on/off) variable. I could check the CL for --verbose somewhere, but I think that's probably overkill. Best regards, Christian Schoenebeck
