[LTP] [PATCH v2 1/2] lib: Add support for test tags

[Li Wang]

On Thu, Oct 24, 2019 at 12:06 AM Cyril Hrubis <chrubis@suse.cz> wrote:

> Hi!
> > Sorry, thanks for the reminder.
> > [I copied the below code to gmail, it always lost the indent]
> >
> > static void print_failure_hints(void)
> > {
> >         unsigned int i;
> >         int hint_printed = 0;
> >         const struct tst_tag *tags = tst_test->tags;
> >
> >         if (!tags)
> >                 return;
> >
> >         for (i = 0; tags[i].name; i++) {
> >                 if (!strcmp(tags[i].name, "linux-git")) {
> >                         if (hint_printed != 1) {
> >                                 hint_printed = 1;
> >                                 print_colored("\nHINT: ");
> >                                 printf("You _MAY_ be missing kernel
> fixes,
> > see:\n\n");
> >                         }
> >
> >                         printf(LINUX_GIT_URL "%s\n", tags[i].value);
> >                 } else if (!strcmp(tags[i].name, "CVE")) {
> >                          if (hint_printed != 2) {
> >                                  hint_printed = 2;
> >                                  print_colored("\nHINT: ");
> >                                  printf("You _MAY_ be vunerable to
> CVE(s),
> > see:\n\n");
> >                           }
> >
> >                         printf(CVE_DB_URL "%s\n", tags[i].value);
>
> This would produce intermixed CVE and linux-git lines unless the tags
> are sorted correctly in the source code, I do not want to depend on the
> order hence the two loops.
>

Yes. But personally I suggest the tags sorted in order. I'm OK with two
loops, it depends on you.


>
> >                 } else {
> >                         print_colored("\nERROR: ");
> >                         printf("tags[%d].name should be linux-git or
> > CVE\n", i);
> >                         return;
>
> I've tried to explain to pvorel already that this is a wrong place to
> assert the tag names. If nothing else this piece of code will be rarely
> triggered and the error would end up ignored.
>

Agree, it seems like too late to see this error in a test fails.


>
> I plan to assert the tag names in the docparse tool that will build the
> test metadata during the LTP build, so that wrong metadata will actually
> fail the LTP build.
>

Cool~ That sounds better.

Reviewed-by: Li Wang <liwang@redhat.com>

-- 
Regards,
Li Wang
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.linux.it/pipermail/ltp/attachments/20191024/bb52de50/attachment-0001.htm>