[Bug tree-optimization/86937] strnlen() of a conditional expression with constant operands not folded

https://gcc.gnu.org/bugzilla/show_bug.cgi?id=86937

--- Comment #2 from GCC Commits <cvs-commit at gcc dot gnu.org> ---
The master branch has been updated by Daniel Barboza <[email protected]>:

https://gcc.gnu.org/g:e7b77e39552d82141f71f039f70885754de72093

commit r17-3337-ge7b77e39552d82141f71f039f70885754de72093
Author: Daniel Barboza <[email protected]>
Date:   Fri Aug 14 18:14:42 2026 -0300

    fold-const-call.cc: add strnlen in fold_const_call [PR86937]

    We're not trying to fold builtin_strnlen calls into constants like we do
    with builtin_strlen.  This results in missed optimizations where we
    have a strnlen that uses a PHI as input:

    const char a[4] = "123";

    int g (int i)
    {
      return __builtin_strnlen (i ? a : "", 4);
    }

    The strnlen call could be folded into a single PHI <0, 3> but instead
    we're calling strnlen with the PHI result:

      # iftmp.1_2 = PHI <&aD.4472(3), ""(2)>
      # VUSE <.MEM_4(D)>
      # RANGE [irange] long unsigned int [0, 4] MASK 0x7 VALUE 0x0
      # USE = nonlocal escaped const-pool { D.4472 } (nonlocal)
      _1 = strnlenD.1862 (iftmp.1_2, 4);
      # RANGE [irange] int [0, 4] MASK 0x7 VALUE 0x0
      _5 = (intD.7) _1;
      # VUSE <.MEM_4(D)>
      return _5;

    After this change:

      # RANGE [irange] int [0, 0][3, 3] MASK 0x3 VALUE 0x0
      # prephitmp_7 = PHI <3(3), i_3(D)(2)>
      # VUSE <.MEM_4(D)>
      return prephitmp_7;

    Bootstrapped and regression tested in x86_64.

            PR tree-optimization/86937

    gcc/ChangeLog:

            * fold-const-call.cc (fold_const_call): fold
            CFN_BUILT_IN_STRNLEN calls.

    gcc/testsuite/ChangeLog:

            * gcc.dg/warn-strnlen-no-nul.c: warnings were
            added/removed/changed due to the new STRNLEN folding.
            * gcc.dg/tree-ssa/pr86937.c: New test.