This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: Wmissing-parameter-type (PR7651 Define -Wextra strictly in terms of other warning flags)
On 19 Dec 2006 21:08:30 -0500, DJ Delorie <dj@redhat.com> wrote:
"Manuel LÃpez-IbÃÃez" <lopezibanez@gmail.com> writes:
> I would say that it is not a new warning but...whatever, OK, I
> withdraw the patch and I will resubmit with its corresponding new
> testcases.
It's not a new warning, but it is a new *option*. You'll need test
cases that test various cases:
1. No option specified
2. Option on (warn), option off (no warn)
3. Option on, marked as an error, marked as a warning overriding -Werror
Really? I was thinking about just 1. and the first part of 2 which is
what I have seen in the testsuite. I don't want to sound pretentious
so, please, believe when I saying that I am asking this sincerely: are
those 3 points a requirement I have overlooked or is it a mere
suggestion? I am a newbie, so I could have easily missed a testcase
that implements point 3.
On the other hand, if this is how new warning options should be
tested, wouldn't it be appropriate to have some kind of automatic
mechanism that applies the different combinations over the same
testcase adjusting the expected output accordingly?
It's the command line processing that you're testing, not just the
code that recognizes the warning in the code.
If that is the goal, I think some mechanism as proposed above will
help a lot. That is, a testcase with a special dg-command like { dg-do
testwarn } { dg-options "-Wmywarning" }. So it tests for the warnings
(the testcase includes the corresponding dg-warning directives), it
also tests that nothing happens without the option and when using
Wno-mywarning (dg-warning directives are converted to dg-bogus), it
can also test for warnings converted to errors and such.
Well, perhaps I am just dreaming since I have no idea how or if this
could be implemented in dejagnu.
Cheers,
Manuel.