On Thu, Mar 3, 2016 at 10:21 AM, David Malcolm <dmalc...@redhat.com> wrote:
> Comment #1 of PR c/68187 identified another overzealous warning
> from -Wmisleading-indentation, with OpenSSL 1.0.1, on this poorly
> indented code:
>
> 115 if (locked)
> 116 i = CRYPTO_add(&e->struct_ref, -1, CRYPTO_LOCK_ENGINE);
> 117 else
> 118 i = --e->struct_ref;
> 119 engine_ref_debug(e, 0, -1)
> 120 if (i > 0)
> 121 return 1;
>
> eng_lib.c:120:9: warning: statement is indented as if it were guarded by...
> [-Wmisleading-indentation]
> if (i > 0)
> ^~
> eng_lib.c:117:5: note: ...this 'else' clause, but it is not
> else
> ^~~~
>
> Line 120 is poorly indented, but the warning is arguably unjustified.
>
> Root cause is that "engine_ref_debug" is actually a debugging macro
> that was empty in the given configuration, so the code effectively
> was:
>
> 117 else // GUARD
> 118 i = --e->struct_ref; // BODY
> 119
> 120 if (i > 0) // NEXT
>
> hence the warning.
>
> But the code as seen by a human is clearly *not* misleading, and
> hence arguably we shouldn't warn for this case.
>
> The following patch fixes this by ruling that if there is non-whitespace
> in a line between the BODY and the NEXT statements, and that this
> non-whitespace is effectively an "unindent" or "outdent" (it's not clear
> to me which of these terms is better), then to not issue a warning.
Cool, this also fixes the false-positives seen in bdwgc, whose coding
style suggests indenting things inside an #ifdef as if it were an
if(), e.g.:
if (a)
foo ();
# ifndef A
bar ();
# endif
...
