On Wed, Sep 12, 2018 at 4:13 PM Nick Desaulniers <ndesaulniers at google.com>
wrote:
> On Wed, Sep 12, 2018 at 1:24 PM Richard Smith <richardsmith at google.com>
> wrote:
> >
> > On Wed, Sep 12, 2018 at 10:38 AM Nick Desaulniers <
> ndesaulniers at google.com> wrote:
> >>
> >> On Tue, Sep 11, 2018 at 5:26 PM Nathan Chancellor
> >> <natechancellor at gmail.com> wrote:
> >> >
> >> > Clang warns if there are missing braces around a subobject
> >> > initializer.
> >> >
> >> > drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c:1447:41: warning: suggest braces
> >> > around initialization of subobject [-Wmissing-braces]
> >> > struct amdgpu_task_info task_info = { 0 };
> >> > ^
> >> > {}
> >> > 1 warning generated.
> >> >
> >> > drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c:262:41: warning: suggest braces
> >> > around initialization of subobject [-Wmissing-braces]
> >> > struct amdgpu_task_info task_info = { 0 };
> >> > ^
> >> > {}
> >> > 1 warning generated.
> >> >
> >> > Reported-by: Nick Desaulniers <ndesaulniers at google.com>
> >> > Signed-off-by: Nathan Chancellor <natechancellor at gmail.com>
> >> > ---
> >> > drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c | 2 +-
> >> > drivers/gpu/drm/amd/amdgpu/gmc_v9_0.c | 2 +-
> >> > 2 files changed, 2 insertions(+), 2 deletions(-)
> >> >
> >> > diff --git a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
> b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
> >> > index 9333109b210d..968cc1b8cdff 100644
> >> > --- a/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
> >> > +++ b/drivers/gpu/drm/amd/amdgpu/gmc_v8_0.c
> >> > @@ -1444,7 +1444,7 @@ static int gmc_v8_0_process_interrupt(struct
> amdgpu_device *adev,
> >> > gmc_v8_0_set_fault_enable_default(adev, false);
> >> >
> >> > if (printk_ratelimit()) {
> >> > - struct amdgpu_task_info task_info = { 0 };
> >> > + struct amdgpu_task_info task_info = { { 0 } };
> >>
> >> Hi Nathan,
> >> Thanks for this patch. I discussed this syntax with our language
> >> lawyers. Turns out, this is not quite correct, as you're now saying
> >> "initialize the first subobject to zero, but not the rest of the
> >> object." -Wmissing-field-initializers would highlight this, but it's
> >> not part of -Wall. It would be more correct to zero initialize the
> >> full struct, including all of its subobjects with `= {};`.
> >
> >
> > Sorry, I think I've caused some confusion here.
> >
> > Elements with an omitted initializer get implicitly zero-initialized. In
> C++, it's idiomatic to write `= {}` to perform aggregate
> zero-initialization, but in C, that's invalid because at least one
> initializer is syntactically required within the braces. As a result, `=
> {0}` is an idiomatic way to perform zero-initialization of an aggregate in
> C.
>
> That doesn't seem to be the case:
> https://godbolt.org/z/TZzfo6 shouldn't Clang warn in the case of bar()?
>
This is a GNU extension. Use -pedantic-errors to turn off extensions, then
Clang and GCC reject bar(): https://godbolt.org/z/pIzI6M
> > Clang intends to suppress the -Wmissing-braces in that case; if the
> warning is still being produced in a recent version of Clang, that's a bug.
> However, the warning suppression was added between Clang 5 and Clang 6, so
> it's very plausible that the compiler being used here is simply older than
> the warning fix.
> >
> > (Long story short: the change here seems fine, but should be unnecessary
> as of Clang 6.)
>
> The warning was identified from clang-8 ToT synced yesterday.
>
Thanks for the testcase. This is a Clang bug. Apparently the warning
suppression only works when the first member is of type 'int'!
https://godbolt.org/z/sxnZvv
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/amd-gfx/attachments/20180913/bd8a6cc1/attachment.html>
