Re: [PATCH v3 05/11] tests/qtest/qos-test: dump environment variables if verbose

[Christian Schoenebeck <qemu_oss@crudebyte.com>]

On Donnerstag, 8. Oktober 2020 14:37:00 CEST Paolo Bonzini wrote:
> On 02/10/20 18:15, Christian Schoenebeck wrote:
> > -int main(int argc, char **argv)
> > +int main(int argc, char **argv, char** envp)
> > 
> >  {
> >  
> >      g_test_init(&argc, &argv, NULL);
> > 
> > +    if (g_test_verbose()) {
> > +        printf("ENVIRONMENT VARIABLES: {\n");
> > +        for (char **env = envp; *env != 0; env++) {
> > +            printf("\t%s\n", *env);
> > +        }
> > +        printf("}\n");
> > +    }
> 
> 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.

> In fact using printf in tests should be forbidden, since glib 2.62 and
> newer _always_ emit TAP.
> 
> Paolo

The reason why I used printf() was because g_test_message() clutters the 
output tremendously, as it always wraps the text of each g_test_message() call 
into:

	"(MSG: %s)\n"

which is inappropriate for multi-line messages as these proposed patches do.

Is that actually a real-life problem? I mean these patches only output 
anything if --verbose CL switch is used, and I don't see any TAP output 
enabled by default. And the --verbose CL switch is usually just used by 
developers for debugging test case issues, isn't it?

Best regards,
Christian Schoenebeck