Re: pahole and gcc-14 issues

[Eduard Zingerman <eddyz87@gmail.com>]

Andrii Nakryiko <andrii.nakryiko@gmail.com> writes:

[...]

> Ok, so sleeping on this a bit more, I'm hesitant to do this more
> generic approach, as now we'll be running a risk of potentially
> looping indefinitely (the max_depth check I added doesn't completely
> prevent this).

Is that different from what happens already?

  /* Check if given two types are identical STRUCT/UNION definitions */
  static bool btf_dedup_identical_structs(struct btf_dedup *d, __u32 id1, __u32 id2)
  {
          [...]
          for (i = 0, n = btf_vlen(t1); i < n; i++, m1++, m2++) {
                  if (...
                      !btf_dedup_identical_structs(d, m1->type, m2->type)) // <-- recursion,
                          return false;                                    //     afaiu it is unbounded.
          }
          return true;
  }

E.g. adding an array to mark visited types and stop descend should stop
inifinite recurusion. I agree with Alan, the patch you shared in the
evening looks interesting. (But it was butchered by gmail,
fixing patches by hand is a bit annoyiong, maybe just attach such files?).

[...]