Semantic of free() and G_free() for G_free_key_value()

[mloskot]

Recently, a bug has been submitted ​to QGIS and forwarded ​to GDAL.

IMHO, the problem is in GRASS.

The function is G_free_key_value​ not proof against nullptr and, given purpose of this function, if pointer to key is null, it should do nothing. This is a standard semantic of free() function in C library. Moreover, G_free function already follows semantic of free(), so G_free_key_value() should do it as well.

In other words, it should read as follows:

void G_free_key_value(struct Key_Value *kv)
{
   if (kv != NULL)
   {
   }
}

[glynn]

Replying to mloskot:

The function is G_free_key_value​ not proof against nullptr and, given purpose of this function, if pointer to key is null, it should do nothing. This is a standard semantic of free() function in C library. Moreover, G_free function already follows semantic of free(), so G_free_key_value() should do it as well.

Fixed in r40399 (trunk).

[hamish]

Replying to glynn:

Fixed in r40399 (trunk).

shall we backport?

[glynn]

Replying to hamish:

shall we backport?

I see no reason not to.

[neteler]

Done in r40403 (6.4) and r40404 (6.5).

[mloskot]

Thank you guys!