Re: [PATCH v4 2/5] t: move reftable/tree_test.c to the unit testing framework

> On Wed, 17 Jul 2024 at 18:09, Karthik Nayak <karthik.188@gmail.com> wrote:
> >
> > Chandra Pratap <chandrapratap3519@gmail.com> writes:
> >
> > > +struct curry {
> > > +     void *last;
> > > +};
> > > +
> > > +static void check_increasing(void *arg, void *key)
> > > +{
> > > +     struct curry *c = arg;
> > > +     if (c->last)
> > > +             check_int(t_compare(c->last, key), <, 0);
> > > +     c->last = key;
> > > +}
> > > +
> > > +static void t_tree(void)
> > > +{
> > > +     struct tree_node *root = NULL;
> > > +     void *values[11] = { 0 };
> >
> > Although we were comparing 'char' above, here we have a 'void *' array.
> > Why?
> 
> The array is passed as a parameter to the 'tree_search()' function which
> requires a void * parameter (i.e. a generic pointer). In the comparison
> function (also passed as a parameter), we cast it to our expected type
> (a character pointer) and then perform the required comparison.

> > > +     struct tree_node *nodes[11] = { 0 };
> > > +     size_t i = 1;
> > > +     struct curry c = { 0 };
> > > +
> > > +     do {
> > > +             nodes[i] = tree_search(values + i, &root, &t_compare, 1);
> > > +             i = (i * 7) % 11;
> >
> > It gets weirder, we calculate 'i' as {7, 5, 2, 3, 10, 4, 6, 9, 8, 1}. We
> > use that to index 'values', but values is '0' initialized, so we always
> > send '0' to tree_search? Doesn't that make this whole thing a moot? Or
> > did I miss something?
> 
> We don't use 'i' to index 'values[]', we use it to calculate the next pointer
> address to be passed to the 'tree_search()' function (the pointer that is 'i'
> ahead of the pointer 'values'), which isn't 0.