<![CDATA[MISRA Discussion Forums - 8.9 Initialization]]> https://forum.misra.org.uk/ Thu, 05 Nov 2026 20:03:05 +0000 MyBB <![CDATA[named vs positional initialization]]> https://forum.misra.org.uk/showthread.php?tid=1751 Thu, 04 Sep 2025 11:27:01 +0000 delirium5223]]> https://forum.misra.org.uk/showthread.php?tid=1751
When mixing them there is a higher chance for human confusion/mistake. e.g. not covering all of them etc



Code:
typedef struct
{
  int x;
  int y;
  int z;
} TypeA;

void test4(void)
{
  TypeA obj = { .y = 2, 3 }; // nok mixed aggregate initialization, named vs positional
  TypeA obj2 = { 3 , .y = 2 }; // nok
  TypeA obj3 = { 3 , 2, 4 }; //ok, not mixed
  TypeA obj4 = { .x = 3 , .z = 4, .y = 2 }; //ok, not mixed
}

Rule 9.6 only prohibits them when they are embedded structs, not for single struct.

Code:
/* Non-compliant - chained designators and implicit positional initializers mixed */
struct T tt = {
1,
.s.x = 2, /* To a human reader, this looks like .z is being initialized */
3, /* tt is actually initialized as { 1, { 2, 3 }, 0 } */
}; /* This also violates Rule 9.2 */
]]>
When mixing them there is a higher chance for human confusion/mistake. e.g. not covering all of them etc



Code:
typedef struct
{
  int x;
  int y;
  int z;
} TypeA;

void test4(void)
{
  TypeA obj = { .y = 2, 3 }; // nok mixed aggregate initialization, named vs positional
  TypeA obj2 = { 3 , .y = 2 }; // nok
  TypeA obj3 = { 3 , 2, 4 }; //ok, not mixed
  TypeA obj4 = { .x = 3 , .z = 4, .y = 2 }; //ok, not mixed
}

Rule 9.6 only prohibits them when they are embedded structs, not for single struct.

Code:
/* Non-compliant - chained designators and implicit positional initializers mixed */
struct T tt = {
1,
.s.x = 2, /* To a human reader, this looks like .z is being initialized */
3, /* tt is actually initialized as { 1, { 2, 3 }, 0 } */
}; /* This also violates Rule 9.2 */
]]> <![CDATA[MISRA C:2012 9.3 clarification]]> https://forum.misra.org.uk/showthread.php?tid=1555 Tue, 24 Nov 2020 09:19:24 +0000 mcks]]> https://forum.misra.org.uk/showthread.php?tid=1555
Is it ok to mix initializers like this, as long as every element gets initialized?

int32_t a[3] = { 1, 2, [2]=3 };

For multi-dimensional arrays, this would be compliant:

int32_t b[2][2] = { { 1, 2 }, { [1]=4 } };

beacuse b[1] is initialized with only a designated initializer.


As for c, both the array subobjects c[0] and c[1] are fully initialized, but there is no initialization of c[2]. Does that make the statement non-compliant?

int32_t c[3][2] = { { 1, 2 }, { 3, 4 } };


What about the arrays below, are they compliant or not, and why? Also, are they violatin 9.2 or 9.4?
int32_t d[1][2] = { [0][1]=2, [0]={1} };
int32_t e[1][2] = { [0]={1}, [0][1]=2 };
int32_t f[3][2] = { [1]={3, 4}, { 5, 6 }, [0][1]=2, [0][0]=1 };
int32_t g[2][2] = { { 1, 2 }, [1][0]=3, 4 };

Let me know what you think. If you have more (tricky) examples, then please share!]]>
Is it ok to mix initializers like this, as long as every element gets initialized?

int32_t a[3] = { 1, 2, [2]=3 };

For multi-dimensional arrays, this would be compliant:

int32_t b[2][2] = { { 1, 2 }, { [1]=4 } };

beacuse b[1] is initialized with only a designated initializer.


As for c, both the array subobjects c[0] and c[1] are fully initialized, but there is no initialization of c[2]. Does that make the statement non-compliant?

int32_t c[3][2] = { { 1, 2 }, { 3, 4 } };


What about the arrays below, are they compliant or not, and why? Also, are they violatin 9.2 or 9.4?
int32_t d[1][2] = { [0][1]=2, [0]={1} };
int32_t e[1][2] = { [0]={1}, [0][1]=2 };
int32_t f[3][2] = { [1]={3, 4}, { 5, 6 }, [0][1]=2, [0][0]=1 };
int32_t g[2][2] = { { 1, 2 }, [1][0]=3, 4 };

Let me know what you think. If you have more (tricky) examples, then please share!]]> <![CDATA[Initialisation of arrays of floats]]> https://forum.misra.org.uk/showthread.php?tid=1357 Wed, 19 Jul 2017 13:04:46 +0000 sdcandy]]> https://forum.misra.org.uk/showthread.php?tid=1357
Code:
uint32_t x[3] = { 0 };

When using floating point types is it also acceptable to use something of the form:

Code:
typedef double float64_t;
float64_t array_a[10] = { 0.0 };

-Andy.]]>
Code:
uint32_t x[3] = { 0 };

When using floating point types is it also acceptable to use something of the form:

Code:
typedef double float64_t;
float64_t array_a[10] = { 0.0 };

-Andy.]]> <![CDATA[Note 9003: could define variable at block scope]]> https://forum.misra.org.uk/showthread.php?tid=1285 Wed, 12 Oct 2016 08:39:58 +0000 PeterW]]> https://forum.misra.org.uk/showthread.php?tid=1285 In my code I have a static const array which is declared outside the block scope of the function where it’s used. Also it is only used by one function, so the Note 9003 is correct.
But the array is large and will ruin the overview of the function. So I want to suppress the warning using /*lint -e9003*/ before the array and /*lint +e9003*/ behind it.
But this has no effect. As I suppress the warning global it works, but this shall not be the way to do.
Any idea how to suppress the warning local for just one variable?]]> In my code I have a static const array which is declared outside the block scope of the function where it’s used. Also it is only used by one function, so the Note 9003 is correct.
But the array is large and will ruin the overview of the function. So I want to suppress the warning using /*lint -e9003*/ before the array and /*lint +e9003*/ behind it.
But this has no effect. As I suppress the warning global it works, but this shall not be the way to do.
Any idea how to suppress the warning local for just one variable?]]> <![CDATA[9.3 and {0}]]> https://forum.misra.org.uk/showthread.php?tid=982 Wed, 11 Sep 2013 19:58:02 +0000 gs]]> https://forum.misra.org.uk/showthread.php?tid=982
Code:
int a[3][2] = {0};
]]>
Code:
int a[3][2] = {0};
]]>