llvm-capstone/clang/test/C/drs/dr4xx.c
Jerin Philip 2e7913342e
[clang] Improve _Alignas on a struct declaration diagnostic (#65638)
Fixes https://github.com/llvm/llvm-project/issues/58637. 

Adds `isAlignas()` method on `AttributeCommonInfo` which accounts for
C++ `alignas` as well as C11 `_Alignas`.

The method is used to improve diagnostic in C when `_Alignas` is used in
C at the wrong location.  This corrects the previously suggested move
of `_Alignas` past the declaration specifier, now warns attribute
`_Alignas` is ignored.

Based on https://reviews.llvm.org/D141177.
2023-11-07 08:27:41 -05:00

375 lines
14 KiB
C

/* RUN: %clang_cc1 -std=c89 -verify=expected,c89only -pedantic -Wno-c11-extensions %s
RUN: %clang_cc1 -std=c99 -verify=expected -pedantic -Wno-c11-extensions %s
RUN: %clang_cc1 -std=c11 -verify=expected -pedantic %s
RUN: %clang_cc1 -std=c17 -verify=expected -pedantic %s
RUN: %clang_cc1 -std=c2x -verify=expected -pedantic %s
*/
/* The following are DRs which do not require tests to demonstrate
* conformance or nonconformance.
*
* WG14 DR401: yes
* "happens before" can not be cyclic
*
* WG14 DR402: yes
* Memory model coherence is not aligned with C++11
*
* WG14 DR404: yes
* Joke fragment remains in a footnote
*
* WG14 DR406: yes
* Visible sequences of side effects are redundant
*
* WG14 DR415: yes
* Missing divide by zero entry in Annex J
*
* WG14 DR417: yes
* Annex J not updated with necessary aligned_alloc entries
*
* WG14 DR419: yes
* Generic Functions
*
* WG14 DR420: yes
* Sytax error in specification of for-statement
*
* WG14 DR425: yes
* No specification for the access to variables with temporary lifetime
*
* WG14 DR434: yes
* Possible defect report: Missing constraint w.r.t. Atomic
*
* WG14 DR435: yes
* Possible defect report: Missing constraint w.r.t. Imaginary
*
* WG14 DR436: yes
* Request for interpretation of C11 6.8.5#6
* Note: This is not really testable because it requires -O1 or higher for LLVM
* to perform its reachability analysis and -Wunreachable-code only verifies
* diagnostic behavior, not runtime behavior. Also, both are a matter of QoI as
* to what they optimize/diagnose. But if someone thinks of a way to test this,
* we can add a test case for it then.
*
* WG14 DR448: yes
* What are the semantics of a # non-directive?
*
* WG14 DR454: yes
* ATOMIC_VAR_INIT (issues 3 and 4)
*
* WG14 DR455: yes
* ATOMIC_VAR_INIT issue 5
*
* WG14 DR459: yes
* atomic_load missing const qualifier
*
* WG14 DR475: yes
* Misleading Atomic library references to atomic types
*
* WG14 DR485: yes
* Problem with the specification of ATOMIC_VAR_INIT
*
* WG14 DR486: yes
* Inconsistent specification for arithmetic on atomic objects
*
* WG14 DR490: yes
* Unwritten Assumptions About if-then
*/
/* WG14 DR412: yes
* #elif
*
* Note: this is testing that #elif behaves the same as #else followed by #if.
*/
#if 1
#elif this is not a valid expression
#else
#if this is not a valid expression
#endif
#endif
/* WG14 DR413: yes
* Initialization
*/
void dr413(void) {
typedef struct {
int k;
int l;
int a[2];
} T;
typedef struct {
int i;
T t;
} S;
/* Ensure that explicit initialization (.t = { ... }) takes precedence over a
* later implicit partial initialization (.t.l = 41). The value should be 42,
* not 0.
*/
_Static_assert((S){ /* c89only-warning {{compound literals are a C99-specific feature}}
expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}}
*/
1,
.t = { /* c89only-warning {{designated initializers are a C99 feature}} */
.l = 43, /* c89only-warning {{designated initializers are a C99 feature}}
expected-note {{previous initialization is here}}
*/
.k = 42,
.a[1] = 19, /* expected-note {{previous initialization is here}} */
.a[0] = 18
},
.t.l = 41, /* expected-warning {{initializer overrides prior initialization of this subobject}} */
.t.a[1] = 17 /* expected-warning {{initializer overrides prior initialization of this subobject}} */
}.t.k == 42, "");
}
/* WG14 DR423: partial
* Defect Report relative to n1570: underspecification for qualified rvalues
*/
/* FIXME: this should pass because the qualifier on the return type should be
* dropped when forming the function type.
*/
const int dr423_const(void);
int dr423_nonconst(void);
_Static_assert(__builtin_types_compatible_p(__typeof__(dr423_const), __typeof__(dr423_nonconst)), "fail"); /* expected-error {{fail}} */
void dr423_func(void) {
const int i = 12;
__typeof__(i) v1 = 12; /* expected-note {{variable 'v1' declared const here}} */
__typeof__((const int)12) v2 = 12;
v1 = 100; /* expected-error {{cannot assign to variable 'v1' with const-qualified type 'typeof (i)' (aka 'const int')}} */
v2 = 100; /* Not an error; the qualifier was stripped. */
}
/* WG14 DR432: yes
* Possible defect report: Is 0.0 required to be a representable value?
*
* We're going to lean on the fpclassify builtin to tell us whether 0.0
* represents the value 0, and we'll test that adding and subtracting 0.0 does
* not change the value, and we'll hope that's enough to validate this DR.
*/
_Static_assert(__builtin_fpclassify(0, 1, 2, 3, 4, 0.0f) == 4, "");
_Static_assert((1.0 / 3.0) + 0.0 == (1.0 / 3.0) - 0.0, ""); /* expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}} */
/* WG14 DR444: partial
* Issues with alignment in C11, part 1
*/
void dr444(void) {
_Alignas(int) int i;
_Alignas(int) struct S {
_Alignas(int) int i;
} s;
/* FIXME: This should be accepted as per this DR. */
int j = (_Alignas(int) int){12}; /* expected-error {{expected expression}} */
_Alignas(int) struct T { /* expected-warning {{'_Alignas' attribute ignored}} */
int i;
};
struct U {
_Alignas(int) int bit : 1; /* expected-error {{'_Alignas' attribute cannot be applied to a bit-field}} */
};
_Alignas(int) typedef int foo; /* expected-error {{'_Alignas' attribute only applies to variables and fields}} */
_Alignas(int) register int bar; /* expected-error {{'_Alignas' attribute cannot be applied to a variable with 'register' storage class}} */
_Alignas(int) void func(void); /* expected-error {{'_Alignas' attribute only applies to variables and fields}} */
/* FIXME: it is correct for us to accept this per 6.7.3p5, but it seems like
* a situation we'd want to diagnose because the alignments are different and
* the user probably doesn't know which one "wins".
*/
_Alignas(int) _Alignas(double) int k;
}
/* WG14 DR447: yes
* Boolean from complex
*
* Ensure that the imaginary part contributes to the conversion to bool, not
* just the real part.
*/
_Static_assert((_Bool)0.0 + 3.0 * (__extension__ 1.0iF), ""); /* c89only-warning {{'_Bool' is a C99 extension}}
expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}}
*/
_Static_assert(!(_Bool)0.0 + 0.0 * (__extension__ 1.0iF), ""); /* c89only-warning {{'_Bool' is a C99 extension}}
expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}}
*/
/* WG14 DR463: yes
* Left-shifting into the sign bit
*
* This DR was NAD and leaves shifting a bit into the high bit of a signed
* integer type undefined behavior, unlike in C++. Note, the diagnostic is also
* issued in C++ for shifting into that bit despite being well-defined because
* the code is questionable and should be validated by the programmer.
*/
void dr463(void) {
(void)(1 << (__CHAR_BIT__ * sizeof(int))); /* expected-warning {{shift count >= width of type}} */
(void)(1 << ((__CHAR_BIT__ * sizeof(int)) - 1));
}
/* WG14 DR478: yes
* Valid uses of the main function
*/
int main(void) {
/* This DR clarifies that C explicitly allows you to call main() in a hosted
* environment; it is not special as it is in C++, so recursive calls are
* fine as well as nonrecursive direct calls.
*/
main(); /* ok */
}
void dr478(void) {
int (*fp)(void) = main; /* ok */
main(); /* ok */
}
/* WG14 DR481: yes
* Controlling expression of _Generic primary expression
*/
void dr481(void) {
/* The controlling expression undergoes lvalue to rvalue conversion, and that
* performs array decay and strips qualifiers.
*/
(void)_Generic("bla", char *: "blu");
(void)_Generic((int const){ 0 }, int: "blu"); /* c89only-warning {{compound literals are a C99-specific feature}} */
(void)_Generic(+(int const){ 0 }, int: "blu"); /* c89only-warning {{compound literals are a C99-specific feature}} */
(void)_Generic("bla", /* expected-error {{controlling expression type 'char *' not compatible with any generic association type}} */
char[4]: "blu"); /* expected-warning {{due to lvalue conversion of the controlling expression, association of type 'char[4]' will never be selected because it is of array type}} */
(void)_Generic((int const){ 0 }, /* expected-error {{controlling expression type 'int' not compatible with any generic association type}}
c89only-warning {{compound literals are a C99-specific feature}}
*/
int const: "blu"); /* expected-warning {{due to lvalue conversion of the controlling expression, association of type 'const int' will never be selected because it is qualified}} */
(void)_Generic(+(int const){ 0 }, /* expected-error {{controlling expression type 'int' not compatible with any generic association type}}
c89only-warning {{compound literals are a C99-specific feature}}
*/
int const: "blu"); /* expected-warning {{due to lvalue conversion of the controlling expression, association of type 'const int' will never be selected because it is qualified}} */
}
/* WG14 DR489: partial
* Integer Constant Expression
*
* The DR is about whether unevaluated operands have to follow the same
* restrictions as the rest of the expression in an ICE, and according to the
* committee, they do.
*/
void dr489(void) {
struct S {
int bit : 12 || 1.0f; /* expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}} */
};
enum E {
Val = 0 && 1.0f /* expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}} */
};
int i;
/* FIXME: mentioning the 'aligned' attribute is confusing, but also, should
* this be folded as an ICE as a GNU extension? GCC does not fold it.
*/
_Alignas(0 ? i++ : 8) char c; /* expected-error {{'aligned' attribute requires integer constant}} */
/* FIXME: this should get the constant folding diagnostic as this is not a
* valid ICE because the floating-point constants are not the immediate
* operand of a cast. It should then also get a diagnostic about trying to
* declare a VLA with static storage duration and the C99 extension warning
* for VLAs in C89.
*/
static int vla[sizeof(1.0f + 1.0f)];
int val[5] = { [1 ? 0 : i--] = 12 }; /* expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}}
c89only-warning {{designated initializers are a C99 feature}}
*/
/* FIXME: this should be the constant folding diagnostic as this is not a
* valid ICE because of the / operator.
*/
_Static_assert(sizeof(0 / 0), "");
/* FIXME: this should also get the constant folding diagnostic as this is not
* a valid ICE because of the = operator.
*/
(void)_Generic(i = 12, int : 0); /* expected-warning {{expression with side effects has no effect in an unevaluated context}} */
switch (i) {
case (int)0.0f: break; /* okay, a valid ICE */
/* FIXME: this should be accepted in C23 and up without a diagnostic, as C23
* added compound literals to the allowed list of things in an ICE. The
* diagnostic is correct for C17 and earlier though.
*/
case (int){ 2 }: break; /* expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}}
c89only-warning {{compound literals are a C99-specific feature}}
*/
case 12 || main(): break; /* expected-warning {{expression is not an integer constant expression; folding it to a constant is a GNU extension}} */
}
}
/* WG14 DR492: yes
* Named Child struct-union with no Member
*/
struct dr492_t {
union U11 { /* expected-warning {{declaration does not declare anything}} */
int m11;
float m12;
};
int m13;
} dr492;
/* WG14 DR496: yes
* offsetof questions
*/
void dr496(void) {
struct A { int n, a [2]; };
struct B { struct A a; };
struct C { struct A a[1]; };
/* Array access & member access expressions are now valid. */
_Static_assert(__builtin_offsetof(struct B, a.n) == 0, "");
/* First int below is for 'n' and the second int is for 'a[0]'; this presumes
* there is no padding involved.
*/
_Static_assert(__builtin_offsetof(struct B, a.a[1]) == sizeof(int) + sizeof(int), "");
/* However, we do not support using the -> operator to access a member, even
* if that would be a valid expression. FIXME: GCC accepts this, perhaps we
* should as well.
*/
(void)__builtin_offsetof(struct C, a->n); /* expected-error {{expected ')'}} \
expected-note {{to match this '('}}
*/
/* The DR asked a question about whether defining a new type within offsetof
* is allowed. C2x N2350 made this explicitly undefined behavior, but GCC and
* Clang both support it as an extension.
*/
(void)__builtin_offsetof(struct S { int a; }, a); /* expected-warning{{defining a type within '__builtin_offsetof' is a Clang extension}} */
}
/* WG14 DR499: yes
* Anonymous structure in union behavior
*/
void dr499(void) {
union U {
struct {
char B1;
char B2;
char B3;
char B4;
};
int word;
} u;
/* Validate that B1, B2, B3, and B4 do not have overlapping storage, only the
* anonymous structure and 'word' overlap.
*/
_Static_assert(__builtin_offsetof(union U, B1) == 0, "");
_Static_assert(__builtin_offsetof(union U, B2) == 1, "");
_Static_assert(__builtin_offsetof(union U, B3) == 2, "");
_Static_assert(__builtin_offsetof(union U, B4) == 3, "");
_Static_assert(__builtin_offsetof(union U, word) == 0, "");
}