[cfe-dev] -fcatch-undefined-behavior false positive with readdir()?

[Richard Smith]

On Wed, Sep 26, 2012 at 12:56 PM, Sean McBride <sean at rogue-research.com>wrote:

> Richard,
>
> It prints 8, for every iteration of the loop.  I guess that's correct
> since the first field of the struct is a uint64_t.


What pointer values does it print? Are they 8 byte aligned?


> Sean
>
>
>
> On Wed, 26 Sep 2012 12:49:38 -0700, Richard Smith said:
>
> >Can you try adding
> >
> >  printf ("%p %zd\\n", d, _Alignof(*d));
> >
> >before the line which fails? Is the pointer correctly aligned?
> >
> >On Wed, Sep 26, 2012 at 12:38 PM, Sean McBride <sean at rogue-research.com
> >wrote:
> >
> >> On Wed, 26 Sep 2012 10:40:47 -0400, Joe Ranieri said:
> >>
> >> >> DIR* dir = opendir("/usr/lib");
> >> >> if (dir)
> >> >> {
> >> >>  struct dirent* d = NULL;
> >> >>  while ((d = readdir(dir)) != NULL)
> >> >>  {
> >> >>   const char* local = d->d_name; // bam!
> >> >>   printf ("%s\n", local);
> >> >>  }
> >> >>  closedir(dir);
> >> >> }
> >> >>
> >> >
> >> > *SNIP*
> >> >
> >> >Just a hunch: is d->d_name an unaligned pointer? We ran into a similar
> >> >issue with gethostbyname.
> >>
> >> Joe,
> >>
> >> Thanks for your reply.  But I'm afraid I don't follow.  Unaligned with
> >> respect to what?  The 'struct dirent' is declared by the system
> basically
> >> like so:
> >>
> >> #pragma pack()
> >> struct dirent {
> >>         uint64_t  d_ino;
> >>         uint64_t  d_seekoff;
> >>         uint16_t  d_reclen;
> >>         uint16_t  d_namlen;
> >>         uint8_t   d_type;
> >>         char      d_name[1024];
> >> }
> >>
> >> Note the 'pragma pack'; is that the alignment you're referring to?  On
> my
> >> system, I don't see any pragma pack with gethostbyname()'s 'struct
> >> hostent', what problem did you have with it?
> >>
> >> I just tried something else: it's not just the d_name field, but
> accessing
> >> any field triggers the SIGILL.
>
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20120926/da50c19e/attachment.html>