[Bug server/26839] Systemtap build failures with clang

["tbaeder at redhat dot com" <sourceware-bugzilla@sourceware.org>]

https://sourceware.org/bugzilla/show_bug.cgi?id=26839

--- Comment #3 from Timm Bäder <tbaeder at redhat dot com> ---
Thanks for the reply, I'll work on this.

I changed the labs() line to:

      unsigned min_score = abs(static_cast<signed>(target.size()) -
static_cast<signed>(it->size()));

since using the unsigned values generates another error:

util.cxx:1545:28: error: call to 'abs' is ambiguous
      unsigned min_score = abs(target.size() - it->size());
                           ^~~
/usr/include/stdlib.h:840:12: note: candidate function
extern int abs (int __x) __THROW __attribute__ ((__const__)) __wur;
           ^
/usr/bin/../lib/gcc/x86_64-redhat-linux/10/../../../../include/c++/10/bits/std_abs.h:56:3:
note: candidate function
  abs(long __i) { return __builtin_labs(__i); }
  ^
/usr/bin/../lib/gcc/x86_64-redhat-linux/10/../../../../include/c++/10/bits/std_abs.h:61:3:
note: candidate function
  abs(long long __x) { return __builtin_llabs (__x); }
  ^
/usr/bin/../lib/gcc/x86_64-redhat-linux/10/../../../../include/c++/10/bits/std_abs.h:71:3:
note: candidate function
  abs(double __x)
  ^
/usr/bin/../lib/gcc/x86_64-redhat-linux/10/../../../../include/c++/10/bits/std_abs.h:75:3:
note: candidate function
  abs(float __x)
  ^
/usr/bin/../lib/gcc/x86_64-redhat-linux/10/../../../../include/c++/10/bits/std_abs.h:79:3:
note: candidate function
  abs(long double __x)
  ^
/usr/bin/../lib/gcc/x86_64-redhat-linux/10/../../../../include/c++/10/bits/std_abs.h:85:3:
note: candidate function
  abs(__GLIBCXX_TYPE_INT_N_0 __x) { return __x >= 0 ? __x : -__x; }
  ^
/usr/bin/../lib/gcc/x86_64-redhat-linux/10/../../../../include/c++/10/bits/std_abs.h:103:3:
note: candidate function
  abs(__float128 __x)
  ^




After adding the 'override' to emit_function() and emit_probe() in
translate.cxx as well as collect_derivation_chain() and
collect_derivation_pp_chain() in elaborate.h, I run into a few other issues:

1) set2 in bpf-bitset.h does not return anything from its assignment operator
but specifies a set2& return value. I fixed this with a 'return *this', but
maybe this is just dead code and should be removed?

2) set1_ref and set1_const_ref have a user-defined copy-assignment operator but
use the implicitly defined copy constructor, which clang warns about:

    ./bpf-bitset.h:108:19: error: definition of implicit copy constructor for
'set1_const_ref' is deprecated because it has a user-declared copy assignment
operator [-Werror,-Wdeprecated-copy]
      set1_const_ref& operator= (const set1_const_ref &);   // not present
                      ^
    ./bpf-bitset.h:256:12: note: in implicit copy constructor for
'bpf::bitset::set1_const_ref' first required here
        return set1_const_ref(data + w2 * i, w2);

Solved this by just defining those missing copy constructors explicitly, e.g.:

+  set1_const_ref(const set1_const_ref &o) : data(o.data), words(o.words) { }



After this, I'm looking at clang and gcc handling -Wformat-nonliteral
differently. Clang warns about this but gcc does not:

common.c:691:20: error: format string is not a string literal
[-Werror,-Wformat-nonliteral]
                vsyslog(LOG_ERR, fmt, va);
                                 ^~~
common.c:693:20: error: format string is not a string literal
[-Werror,-Wformat-nonliteral]
                vfprintf(stderr, fmt, va);
                                 ^~~

And another time in stapbpf.cxx:

stapbpf.cxx:257:20: error: format string is not a string literal
[-Werror,-Wformat-nonliteral]
  vfprintf(stderr, str, va);


Not quite sure how to best handle these. Adding a

#pragma GCC diagnostic push
#pragma GCC diagnostic ignored "-Wformat-nonliteral"
// Code
#pragma GCC diagnostic pop

block works for clang as well and is also done in e.g. stapbpf/bpfinterp.cxx
for other code. What do you think?


Other than that, the only problem left is the printsig() stuff from above.

-- 
You are receiving this mail because:
You are the assignee for the bug.