Initializer element is not constant

Hi,

It has been a LONG TIME since I've done any programming in C. I have two old programs still in use which were developed on msdos, using the "Mark Williams Let's C" compiler. We're talking 1980's stuff here.

Anyway, I thought to try to port them to linux. Can anyone tell me what I'm doing wrong?

Here are the relevant portions of indatax.c:

char *namectrl = NULL;
char *name1 = NULL;
char *name2 = NULL;

typedef struct {
char *var; /* destination for storage of the data */
size_t len; /* max len of the data */
char *(*xlat)(); /* translation routine */
} DATUM;

DATUM recip_data[] = {

{  namectrl, NAMECTRL_LEN, make_upper },
{  name1, NAME1_LEN, make_upper },
{  name2, NAME2_LEN, make_upper },

[...]

And these are the errors...

indatax.c:124: initializer element is not constant indatax.c:124: (near initialization for recip_data[0].var') indatax.c:125: initializer element is not constant indatax.c:125: (near initialization for recip_data[1].var') indatax.c:126: initializer element is not constant indatax.c:126: (near initialization for `recip_data[2].var') [...]

gcc version 2.95.3

Thanks,

Scott Swanson
Pendroy, Montana

Hi,

can u point us to line 124??

I've encountered a similar error message before when i using HP-UX on Intel Itanium. The compiler complained that i used a non-constant in a static array declaration... char a[ maxLen ] ;

in the end, I had to malloc/calloc them... (since i didnt know how to tweak or instruct the compiler to ignore such errors... since the code was already old and compilable on other hp-ux platforms)

-Ron

have u tried changing the first field of each element?? something like... replacing those char ptrs with something static?

example:

DATUM recip_data[] = {
{ "i am a constant str", NAMECTRL_LEN, make_upper }, { "i am a constant str", NAME1_LEN, make_upper },

u might notice that some of "initializer element is not constant" errors will be significantly reduced....

perhaps, the compiler is complaining that u have used something "dynamic"(or at least non-static) in defining a "statically" allocated data structure(recip_data)
=(

perhaps u can change the way u define/declare ur variables/data structures....

As I recalled, the ability to HAVE dynamic pointers was one of the primary benefits of using c in the first place! Later in the file I have this routine which allocates space for each pointer in the structure:

alloc_space(r) RECORD *r;

{	register size_t i;
	register char **v;
	
	for(i=0; i<r->n; i++)
	{	if(r->data[i].len == 0 || r->data[i].len == 0xFFFF) continue;
		v = (r->data[i].var);
		if (*v == NULL)
		{	*v = (char *) calloc(1,r->data[i].len + 1);
			memset(*v,'A'+i,r->data[i].len);
		}
	}
}

I guess i just don't understand why a "modern" compiler wants to restrict my options in this fashion....

but u can.... =)
have u tried dynamically allocating space for recip_data?? instead of the usual static declaration? char a[10] <-- this is static
char a[]= {'1','2','3','4'} <-- this is static with definition and declartion in one char *a = calloc.....

have u tried it? using malloc/calloc for defining DATUM recip_data??

im not good with explanations....
but i bet the gurus like JBG could explain the solution/scenario alot better. =)

Well yes, I guess I could [m|c]alloc the recip_data array. But how do I then populate it?

recip_data[0].var = namectrl;
recip_data[0].len = NAMECTRL_LEN;
recip_data[0].xlat = make_upper;
etc...

How tedious is THAT!? Especially considering that the code has eight such arrays, averaging 6 elements per array.

Maybe I'm not understanding your suggestion. As I mentioned, it has been quite a while since I did any c programming. I just find it strange that code which compiled fine 20 years ago now will not. I could understand if it just generated warnings....

Hi Scott,

namectrl is NULL, so why would you want to do this? This is just like writing

recip_data[0] = NULL;

I am not quite sure what you intend to do, but I guess that you want namectrl to change when you change recip_data[0]. To accomplish this, you could make recip_data[0] a pointer to a pointer:

char *namectrl = NULL;
char *name1 = NULL;
char *name2 = NULL;

typedef struct {
char **var; /* destination for storage of the data */
size_t len; /* max len of the data */
char *(*xlat)(); /* translation routine */
} DATUM;

DATUM recip_data[] = {

{  &namectrl, NAMECTRL_LEN, make_upper },
{  &name1, NAME1_LEN, make_upper },
{  &name2, NAME2_LEN, make_upper },

[...]

Now you can change namectrl by saying
*recip_data[0].var = "test string";

Hope this helps,
Christoph

``There's no dark side of the moon, really
Matter of fact, it's all dark''

--Pink Floyd

Right! As I said, my c is quite rusty. Now as I recall, there was some problem with the old dos compiler which prevented this syntax but allowed the erroneous one. That, or I didn't understand what I was doing and the compiler let me get away with it.... Looking at the code in more detail, I see that it relies upon some bcd math module I'd written in assembler, so the odds of this getting ported are pretty slim. Thanks for the help, though!

Scott Swanson