Re: [PATCH] careadlinkat: fix GCC 10 workaround

[Bernhard Voelker]

Hi Paul,

On 2020-05-11 19:47, Paul Eggert wrote:
> +2020-05-11  Paul Eggert  <eggert@cs.ucla.edu>
> +
> +     careadlinkat: fix GCC 10 workaround
> +     * lib/careadlinkat.c (careadlinkat) [GCC_LINT]:
> +     Massage the code so that it’s closer to what it was before
> +     the GCC 10.1.0 workaround was introduced.  This fixes
> +     a loop when !buffer and the bug workaround is in effect.
> +     Remove unnecessary casts.  Defend in a different way
> +     against (buffer && !buffer_size), by adding at least 1
> +     to buf_size each time through the loop.

GCC-10.2.1 (which I have here on Tumbleweed) still complains:

  $ ./gnulib-tool --create-testdir --dir=/tmp/testdir careadlinkat
  $ cd /tmp/testdir && ./configure && make
  ...
  gcc -DHAVE_CONFIG_H -I. -I..  -DGNULIB_STRICT_CHECKING=1   -g -O2 -MT 
careadlinkat.o -MD -MP -MF \
    .deps/careadlinkat.Tpo -c -o careadlinkat.o careadlinkat.c
  careadlinkat.c: In function 'careadlinkat':
  cc1: warning: function may return address of local variable 
[-Wreturn-local-addr]
  careadlinkat.c:73:8: note: declared here
     73 |   char stack_buf[1024];
        |        ^~~~~~~~~

And there doesn't seem to be any activity on the GCC bug and related ones.
  https://gcc.gnu.org/bugzilla/show_bug.cgi?id=95044

Currently, we can only get avoid the warning via GCC_LINT.
Do you have another idea how to avoid it?  E.g. referencing
static memory instead of stack memory?

Thanks & have a nice day,
Berny