JsonLogTask: fix/improve value comparisons

[ralphlange]

fixes #43

[simon-ess]

To be clear, this comment is just pointing out that this will still return some false negetives and print a log, depending on the junk bytes behind the null. So this is an improvement but not perfect?

[ralphlange]

Correct.
Doing min(strlen(), MAX_LENGTH) for every element seemed excessive.

[mdavidsaver]

Is there some guarantee that bytes following a terminating nil will also be nil? eg. does "foo\0blah" equal "foo\0\0\0\0\0"? memcmp() will say "no", strncmp() will say "yes".

Also, the preceding macro and switch statement could be mostly replaced by use of dbValueSize().

Something like:

size_t size = pLogData->is_array ? pLogData->old_log_size : 1;
if(pLogData->type==DBR_STRING) {
    for(size_t i=0; i<size; i++) {
        if(strncmp(pa->a_string[i], pb->a_string[i], MAX_STRING_SIZE)!=0) {
            return 1;
        }
    }
} else {
    return memcmp(pa, pb, size*dbValueSize(pLogData->type))!=0;
}
return 0;

[ralphlange]

Is there some guarantee that bytes following a terminating nil will also be nil? eg. does "foo\0blah" equal "foo\0\0\0\0\0"? memcmp() will say "no", strncmp() will say "yes".

No, there isn't. (See above.)
I found doing multiple memcmp/strncmp excessive, maybe it's not.
Given that arrays of strings are pretty rare...

Also, the preceding macro and switch statement could be mostly replaced by use of dbValueSize().

Something like:

size_t size = pLogData->is_array ? pLogData->old_log_size : 1;
if(pLogData->type==DBR_STRING) {
    for(size_t i=0; i<size; i++) {
        if(strncmp(pa->a_string[i], pb->a_string[i], MAX_STRING_SIZE)!=0)) {
            return 1;
        }
    }
} else {
    return memcmp(pa, pb, size*dbValueSize(pLogData->type))!=0;
}
return 0;

That's clearly beyond the original scope of this fix, but very reasonable.

[mdavidsaver]

No, there isn't. (See above.)

Ah, now that exchange registers with me.

imo. there seems almost no point to memcmp() past the first terminating nil. The chances of a successful equality seem vanishing small ($\frac{1}{256^N}$), statistically similar to just return 1.

[anjohnson]

Much simpler, LGTM!

[ralphlange]

Superseded by #50