Re: [patch] Use inttypes macros for size_t format string

[Andreas Schneider <asn@xxxxxxxxxxxxxx>]

On Wednesday, 15 January 2020 16:25:44 CET g4-lisz@xxxxxxxxxxxx wrote:
> On 15.01.20 14:35, Jakub Jelen wrote:
> > On Wed, 2020-01-15 at 13:01 +0100, g4-lisz@xxxxxxxxxxxx wrote:
> >> Hi there, here's a patch for fixing a printf format string issue when
> >> compiling with MinGW (and possibly other "architectures"...).
> >> 
> >> A big thanks to Zdenek OGAR Skalak for the hint!
> > 
> > Looks good to me. Grepping through the rest of the code shows that
> > there will most probably be more issues like this:
> > 
> > $ git grep "PRIdS" | wc -l
> > 10
> > $ git grep "%zu" | wc -l
> > 27
> > 
> > Could you check also the other cases to make sure we can address it in
> > the whole codebase?
> 
> Actually, `size_t` is unsigned. So IMHO `PRIuS` should be the format
> macro...
> 
> But in priv.h we have only
> 
>    #define PRIdS "zd"
> 
> or
> 
>   #define PRIdS "Id"
> 
> for WIN32
> 
> So maybe we should replace this with `PRIuS` and `zu` / `Iu`?
> 
> This would at least solve the situation in dh-gex.c:408 where we have:
> 
>         rc = fscanf(moduli,
>                     "%31s %zu %zu %zu %zu %31s %4095s\n",
>                     timestamp,
>                     &type,
>                     &tests,
>                     &tries,
>                     &size,
>                     generator,
>                     modulus);
> 
> and type, tests, tries and size are of size_t. Or we have to change them
> to ssize_t, but then also the `best_size` argument of
> ssh_retrieve_dhgroup_file() had to be signed.
> 
> What you think?

For size_t we should use %zu and for ssize_t %zd which is ANSI C99.

See https://en.cppreference.com/w/c/io/fprintf


	Andreas

-- 
Andreas Schneider                 asn@xxxxxxxxxxxxxx
GPG-ID:     8DFF53E18F2ABC8D8F3C92237EE0FC4DCC014E3D