zero-sized array in struct/union
-
I have a legacy C code:
char* file_name[0];
which generate a warning:
warning C4200: nonstandard extension used : zero-sized array in struct/unionit is correct to write:char* file_name[_MAX_PATH];
? I don't know the impact of this modify...
Read carefully the documentation[^] (see the sample code). Using
_MAX_PATH(or whatever>0) is correct, the impact is in memory: each time thestructis allocated,_MAX_PATHcharacter pointers are allocated too. You might instead choose to disable the warning, if it makes sense (e.g. there is an additional field in the struct specifying the actual size of the array). -
_Flaviu wrote:
Or
char* file_name[1];
Yes. The only reason someone would declare zero-length is to dynamic allocate the array. You should change the array length to [1]. If you change it to _MAX_PATH (260) then you will be wasting 1036 bytes on a 32 bit machine and wasting 2072 bytes on a 64 bit machine. Wasting bytes is punishable by death. Best Wishes, -David Delaune
-
_Flaviu wrote:
Or
char* file_name[1];
Yes. The only reason someone would declare zero-length is to dynamic allocate the array. You should change the array length to [1]. If you change it to _MAX_PATH (260) then you will be wasting 1036 bytes on a 32 bit machine and wasting 2072 bytes on a 64 bit machine. Wasting bytes is punishable by death. Best Wishes, -David Delaune
-
Read carefully the documentation[^] (see the sample code). Using
_MAX_PATH(or whatever>0) is correct, the impact is in memory: each time thestructis allocated,_MAX_PATHcharacter pointers are allocated too. You might instead choose to disable the warning, if it makes sense (e.g. there is an additional field in the struct specifying the actual size of the array). -
Quote:
You should change the array length to [1]
Quote:
Wasting bytes is punishable by death
Take your own conclusions. :-D
Hmmm, The law states that wasting bytes less or equal to 1 * sizeof(pointer) is allowed but only in the month of August. I guess he could remove the array qualifier but then that would probably break his compile. :) Best Wishes, -David Delaune
-
The original code is:
typedef struct { .... .... char\* file\_name\[0\]; /\* File name in Unicode. \*/ }; // warning C4094: untagged 'struct' declared no symbolsalso, I get another warning here:
warning C4094: untagged 'struct' declared no symbolsI don't know how to get rid of this warnings ...You are using a typedef but have not given it the name that you wish to use. It should be something like:
typedef struct {
....
....
char* file_name[0]; /* File name in Unicode. */
} myStruct;
// myStruct is now an alias for the above structureAlso the comment on the last line makes no sense; firstly it is declaring an aray of pointers rather than characters. And secondly, you should not store Unicode characters in a
chartype array. It will most likely cause problems at run time. The zero length array is possibly valid, but it depends on how the code uses the struct. It can be used as a placeholder name for space that will be allocated for a dynamic structure at run time. Something like:struct foo
{
int i;
char text[0];
};// ... other code
struct foo* myFoo = (struct foo*)malloc(sizeof(struct foo) + 20); // additional 20 bytes for the char data.
-
I guess disabling this warning is best solution … how can I do that ? With pragma statement ? If yes, which version of pragma should I use ?
-
Read carefully the documentation[^] (see the sample code). Using
_MAX_PATH(or whatever>0) is correct, the impact is in memory: each time thestructis allocated,_MAX_PATHcharacter pointers are allocated too. You might instead choose to disable the warning, if it makes sense (e.g. there is an additional field in the struct specifying the actual size of the array). -
Hmmm, The law states that wasting bytes less or equal to 1 * sizeof(pointer) is allowed but only in the month of August. I guess he could remove the array qualifier but then that would probably break his compile. :) Best Wishes, -David Delaune
-
_Flaviu wrote:
Or
char* file_name[1];
Yes. The only reason someone would declare zero-length is to dynamic allocate the array. You should change the array length to [1]. If you change it to _MAX_PATH (260) then you will be wasting 1036 bytes on a 32 bit machine and wasting 2072 bytes on a 64 bit machine. Wasting bytes is punishable by death. Best Wishes, -David Delaune
-
Yes, but that is punished by [tar and feathering](https://en.wikipedia.org/wiki/Tarring\_and\_feathering) :laugh: Best Wishes, -David Delaune
-
Quote:
_MAX_PATH (260) then you will be wasting 236 bytes on a 32 bit machine and wasting 472 bytes on a 64 bit machine
Hey David, the math there is not clear to me. Do I need more caffeine this morning?
-
Quote:
_MAX_PATH (260) then you will be wasting 236 bytes on a 32 bit machine and wasting 472 bytes on a 64 bit machine
Hey David, the math there is not clear to me. Do I need more caffeine this morning?
If you are on C11 ... C11 6.7.9/14 allows the option
char file_name[];
It was addedd for exactly that reason
In vino veritas
-
If you are on C11 ... C11 6.7.9/14 allows the option
char file_name[];
It was addedd for exactly that reason
In vino veritas
-
The original code is:
typedef struct { .... .... char\* file\_name\[0\]; /\* File name in Unicode. \*/ }; // warning C4094: untagged 'struct' declared no symbolsalso, I get another warning here:
warning C4094: untagged 'struct' declared no symbolsI don't know how to get rid of this warnings ...pretty sure you could just use char* name; since the array size is 0.
-
You are using a typedef but have not given it the name that you wish to use. It should be something like:
typedef struct {
....
....
char* file_name[0]; /* File name in Unicode. */
} myStruct;
// myStruct is now an alias for the above structureAlso the comment on the last line makes no sense; firstly it is declaring an aray of pointers rather than characters. And secondly, you should not store Unicode characters in a
chartype array. It will most likely cause problems at run time. The zero length array is possibly valid, but it depends on how the code uses the struct. It can be used as a placeholder name for space that will be allocated for a dynamic structure at run time. Something like:struct foo
{
int i;
char text[0];
};// ... other code
struct foo* myFoo = (struct foo*)malloc(sizeof(struct foo) + 20); // additional 20 bytes for the char data.
I was about to write an answer when i saw this. Yes, zero length char arrays at the end of a struct appeared to be quite common in C programming some years (or decades, rather) ago. I haven't seen it in any C**++** code ever, although it probably works the same. Whatever you wish to achieve, there's probably a better solution available in C++ syntax. Usually, std::string is the go to solution here. That said, yes, it must be
char [], notchar* [], otherwise it doesn't make any sense at all.GOTOs are a bit like wire coat hangers: they tend to breed in the darkness, such that where there once were few, eventually there are many, and the program's architecture collapses beneath them. (Fran Poretto)
-
I guess disabling this warning is best solution … how can I do that ? With pragma statement ? If yes, which version of pragma should I use ?
Don't disable warnings unless you are 100% sure what they're telling you, 100% sure that this is not a problem for the syntactic and semantic functionality of your code, and at least 90% sure there's no reasonable way to avoid them. Under these conditions, the best way is to use #pragma push immediately before the disable command and #pragma pop after the code that causes the warning. That way you can be sure that the remainder of the code will use the same warning settings as defined in the compiler options.
GOTOs are a bit like wire coat hangers: they tend to breed in the darkness, such that where there once were few, eventually there are many, and the program's architecture collapses beneath them. (Fran Poretto)
-
I was about to write an answer when i saw this. Yes, zero length char arrays at the end of a struct appeared to be quite common in C programming some years (or decades, rather) ago. I haven't seen it in any C**++** code ever, although it probably works the same. Whatever you wish to achieve, there's probably a better solution available in C++ syntax. Usually, std::string is the go to solution here. That said, yes, it must be
char [], notchar* [], otherwise it doesn't make any sense at all.GOTOs are a bit like wire coat hangers: they tend to breed in the darkness, such that where there once were few, eventually there are many, and the program's architecture collapses beneath them. (Fran Poretto)
