[PATCH] D51211: [Sema] Emit -Wformat properly for bitfield promotions.

[Aaron Ballman via Phabricator via cfe-commits]

aaron.ballman accepted this revision.
aaron.ballman added a comment.
This revision is now accepted and ready to land.

LGTM!



================
Comment at: test/Sema/format-strings-bitfield-promotion.c:1
+// RUN: %clang_cc1 -triple x86_64-unknown-unknown -fsyntax-only -verify %s
+
----------------
ebevhan wrote:
> aaron.ballman wrote:
> > ebevhan wrote:
> > > ebevhan wrote:
> > > > aaron.ballman wrote:
> > > > > aaron.ballman wrote:
> > > > > > Running your test through GCC looks like the behavior matches here for C; can you also add a C++ test that demonstrates the behavior does not change?
> > > > > > 
> > > > > > https://godbolt.org/z/zRYDMG
> > > > > Strangely, the above godbolt link dropped the output windows, here's a different link that shows the behavioral differences between C and C++ mode in GCC: https://godbolt.org/z/R3zRHe
> > > > Hmm, I'll have a look at this.
> > > That gcc godbolt is a bit odd. The type of the bitfield expression in the C++ example is `long` and not `int`, but in Clang, it's clearly being converted. If I change the example a bit, we get this warning:
> > > ```
> > > <source>:11:12: warning: format '%d' expects argument of type 'int', but argument 2 has type 'long int' [-Wformat=]
> > >    11 |   printf("%d", bf.a); // expected-warning {{format specifies type 'long' but the argument has type 'int'}}
> > >       |           ~^   ~~~~
> > >       |            |      |
> > >       |            int    long int
> > > ```
> > > But in Clang, we get a cast to `int`:
> > > ```
> > >     | `-ImplicitCastExpr 0xd190748 <col:17, col:20> 'int' <IntegralCast>
> > >     |   `-ImplicitCastExpr 0xd190730 <col:17, col:20> 'long' <LValueToRValue>
> > >     |     `-MemberExpr 0xd190618 <col:17, col:20> 'long' lvalue bitfield .a 0xd18f790
> > >     |       `-DeclRefExpr 0xd1905f8 <col:17> 'struct bitfields':'bitfields' lvalue Var 0xd18fa18 'bf' 'struct bitfields':'bitfields'
> > > ```
> > > 
> > > So gcc and Clang are doing things differently here.
> > > 
> > > The code in `isPromotableBitField` says:
> > > ```
> > >   // FIXME: C does not permit promotion of a 'long : 3' bitfield to int.
> > >   //        We perform that promotion here to match GCC and C++.
> > > ```
> > > but clearly gcc isn't doing this in the C++ case. The comments also mention some things about gcc bugs that Clang does not follow, but that's in reference to a C DR.
> > C++ disallows the rank conversion from int to long as well. [conv.prom]p1 does not apply because `long int` has a higher rank than `int`, but [conv.prom]p5 allows the promotion if the range of values is identical between the two types.
> > 
> > C makes this UB in several ways -- you can't have a bit-field whose type is something other than int, unsigned int, or _Bool (6.7.2.1p5) or promoting from types other than those (6.3.1.1p2), but otherwise matches the C++ behavior in terms of promotion (including the rank conversion).
> > 
> > You may have to dig further into what Clang is doing, but I would guess that the diagnostics should be triggered in both C and C++ similarly.
> > 
> > Ultimately, I'd like to see tests for cases where `sizeof(int) == sizeof(long)`, `sizeof(int) != sizeof(long)`, and variants for C and C++ of each.
> I'm not sure the warning should trigger in C++; the behavior is correct there. The expression in those cases should be of type `long`, not `int`. The bitfield promotions in C++ say that values _can_ be promoted if the value fits in `int`, but the rules in C say that the value _is_ promoted.
> 
> The strange promotion behavior only occurs in C because of the issue with bitfields larger than int. It's not really permitted according to the standard, but it's supported anyway to match C++. Though, it ends up not matching C++ due to these promotion differences.
> 
> I'll add tests for the different int and long sizes, though the only case where it would make a difference would be if int was larger than 32 bits, which it isn't on any target.
> The bitfield promotions in C++ say that values _can_ be promoted if the value fits in int, but the rules in C say that the value _is_ promoted.

Ahhh, that explains the differences (my eyes glossed over that in the standards text), thank you!


CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D51211/new/

https://reviews.llvm.org/D51211