[PATCH] D66043: Add to -Wparentheses case of bitwise-and ("&") and bitwise-or ("|") verses conditional operator ("?:")
Richard Trieu via Phabricator via cfe-commits
cfe-commits at lists.llvm.org
Mon Aug 12 19:38:02 PDT 2019
rtrieu marked an inline comment as done.
rtrieu added inline comments.
Comment at: test/Sema/parentheses.c:156
+ (void)(x ^ b ? 1 : 2); // no warning, ^ is often used as logical xor
+ (void)(x || b ? 1 : 2); // no warning, logical operator
> rtrieu wrote:
> > jfb wrote:
> > > rtrieu wrote:
> > > > jfb wrote:
> > > > > I don't understand why `^` is different. What do you mean by "often used as a logical xor`?
> > > > In C++, there's the bitwise operators |, &, and ^. And there's the logical operators || and &&. Since there's no ^^ for a logical-xor, many people will just use the bitwise-xor ^ instead. Since this isn't warning on logical operators, it makes sense to exclude the bitwise-xor that is often used as logical-xor.
> > > So code is often buggy when it uses `|` and `&` as diagnosed by this patch, but is rarely buggy when it uses `^`?
> > That's correct. From my testing, &&, || and ^ all had low bug finding rates and didn't make sense to include into this warning while | and & had a high bug finding rate.
> OK thanks for clarifying. Could you explain this instead of "logical xor"? It seems useful to know when reading your code that in your experience `^` wasn't a source of bugs as much as the others.
I have added comments for this test and for the code where xor is excluded.
CHANGES SINCE LAST ACTION
More information about the cfe-commits