On Mon, Jul 26, 2021 at 01:27:48PM +0100, Bruce Richardson wrote:
> On Mon, Jul 26, 2021 at 01:16:27PM +0100, John Levon wrote:
> > meson test was not capturing the intended output from the child
> > process; force a flush to ensure it reaches the test log.
> >
> > Signed-off-by: John Levon <john.le...@nutanix.com>
> > ---
> > app/test/process.h | 1 +
> > 1 file changed, 1 insertion(+)
> >
> > diff --git a/app/test/process.h b/app/test/process.h
> > index a09a088477..0ed91a939e 100644
> > --- a/app/test/process.h
> > +++ b/app/test/process.h
> > @@ -110,6 +110,7 @@ process_dup(const char *const argv[], int numargs,
> > const char *env_value)
> > for (i = 0; i < num; i++)
> > printf("'%s' ", argv_cpy[i]);
> > printf("\n");
> > + fflush(stdout);
> >
> I thought that "\n" normally flushes in most cases, but if explicit flush
> is necessary for some cases, let's add it.
stdout is only line-buffered if it's to a tty, but regardless, meson test
appears to wrap all output into tmp files. I didn't dig into exactly what's
going on with meson (or why the fork() matters), but certainly you don't get
this output *at all* under meson test (even with -v), but you do running
dpdk_test directly.
Furthermore, looks like others have found the same thing as there are several
other direct fflush()es.
regards
john
