json_object_get_int does not set EINVAL on invalid string

[fedefrancescon]

Describe the bug
The doc states that json_object_get_int should set errno to EINVAL.
This is not the case! Looking at the code, in case the type is string and the parsing fails it returns 0 but errno is not set.

Steps To Reproduce
Compile the following code with no particular flags or settings:

#include <assert.h>
#include <json-c/json.h>
#include <stdio.h>
#include <errno.h>

int main(const int argc, const char *argv[]){
  const char vals[][10] = {
    "b",
    "0xb",
    "0xB",
    "0xB0",
    "0B",
    "pippo",
    "pluto"
  };

  for (size_t i = 0; i < sizeof(vals)/sizeof(vals[0]); i++) {
    struct json_object *jtmp = json_object_new_string(vals[i]);
    errno = 0;
    int r = json_object_get_int(jtmp);
    printf("s: %-10s => i: %3lld errno: %d\n", vals[i], r, errno);
    json_object_put(jtmp);
  }

  return 0;
}

Will give you that output:

s: b          => i:   0 errno: 0
s: 0xb        => i:   0 errno: 0
s: 0xB        => i:   0 errno: 0
s: 0xB0       => i:   0 errno: 0
s: 0B         => i:   0 errno: 0
s: pippo      => i:   0 errno: 0
s: pluto      => i:   0 errno: 0

Version and Platform

• json-c version: 777dd06 , 0.15
• OS: Arch Linux x64 , Kirkstone on aarch64

How to Proceed?
I was on my way to issue a PR to fix this (with related test), it's a really easy fix. But I noticed that there is some inconsistency within:

• json_object_get_double
• json_object_get_int
• json_object_get_int64
• json_object_get_uint64

int and double both state that invalid conversions should set errno = EINVAL, double actually seems to do this!
int64 and uint64 do not state the same and simply (at first glance) returns 0 in case of invalid parsing.

So, before going on an pushing a PR. Which way should I go? Which is the indended behaviour?
Could I go on and fix the missing errno = EINVAL or just remove the errno setting statement from the doc?

[hawicz]

The fix for this belongs in the json_parse_int64() function. That calls strtoll() with the expectation that it will set errno on failure, but the man page states that this is optional ("...may also set..."):

The implementation may also set errno to EINVAL in case no conversion was performed (no digits seen, and 0 returned).

so we'll need to set it explicitly to get consistent behavior.

[hawicz]

Note that the errno: 0 result for the cases that start with an actual "0" are the intended behavior. In those, when the string is "coerced to a int" (as per the docs) the answer is 0 and the extra characters are supposed to be ignored.