This is the mail archive of the libc-alpha@sourceware.org mailing list for the glibc project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: [RFC v5 01/21] sunrpc/clnt_udp: Ensure total_deadline is initalised


On Thu, Sep 5, 2019 at 6:31 PM Jeff Law <law@redhat.com> wrote:
>
> On 9/5/19 6:46 PM, Alistair Francis wrote:
> > On Thu, Sep 5, 2019 at 8:02 AM Joseph Myers <joseph@codesourcery.com> wrote:
> >>
> >> On Thu, 5 Sep 2019, Alistair Francis wrote:
> >>
> >>> On Thu, Aug 29, 2019 at 10:34 AM Zack Weinberg <zackw@panix.com> wrote:
> >>>>
> >>>> On Thu, Aug 29, 2019 at 1:22 PM Joseph Myers <joseph@codesourcery.com> wrote:
> >>>>> On Thu, 29 Aug 2019, Alistair Francis wrote:
> >>>>> The long pole is definitely the ml2014 build environment, unless for some reason we need the new version of pip first? I don't actually know.  I'm assu
> >>>>>> Even though total_deadline won't be accessed uninitalised GCC can still
> >>>>>> complain that it is accessed unitalised, to avod those errors let's make
> >>>>>> sure we initalise it to 0.
> >>>>>
> >>>>> It's glibc practice (although missing from
> >>>>> <https://sourceware.org/glibc/wiki/Style_and_Conventions>) that we *don't*
> >>>>> add initializations like that to avoid warnings.
> >>>>
> >>>> Although this has historically been glibc practice, I think it is
> >>>> unwisely incautious, and we should change the policy to be that we
> >>>> *do* add initializations whenever the compiler thinks a variable even
> >>>> _might_ be used uninitialized.
> >>>
> >>> Does that mean this patch is ok?
> >>
> >> No.  You can't deduce consensus like that from two different views on a
> >> patch or a convention.  Even if we were to change the convention regarding
> >> how to silence such warnings, I see reason to have any less requirement
> >> for comments explaining why the warning is a false positive and that the
> >> initializer is only there to silence a warning than there is for the
> >> DIAG_* macros.
> >
> > No worries, I'll happily change the patch, I just want to make sure I
> > change it to the right thing.
> >
> > I'll:
> >  - Investigate filing a GCC bug for this false positive
> All we need is the .i file (Add -save-temps to the compilation line), a
> copy of the full compilation line and the target triplet.  No need to
> try and create a minimal testcase or anything like that.

Thanks for that info. I have filed the bug:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=91691

I'll send this patch out separately and we can go from there.

Alistair

>
> Jeff


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]