[mborland]

il_display prints a pointer to an entity outside the file-scope memory
region with its raw address. One example is the name of a function's
parameter, in the param-type entry of the function's type:

name:                    **possible non-file-scope ptr** (0x7c74a620a0): "x"

The address changes from run to run. The test output filter
(dev_tools/pylibs/edgtest/filter/__init__.py, _REGEX_REPLACEMENTS)
normalizes the other addresses (@…, @0x…, .text+0x…) but not this one.
So the recorded output of any test that displays such an IL never matches
twice.

[wchilders-nvidia]

I wonder if we should consider making a change to il_display instead @daveedvdv-nvidia @nv-cmeerw thoughts ?

[daveedvdv-nvidia]

An il_display.c change might be reasonable here. I'm seeing:

    if (displaying_file_scope_il && !is_file_scope_entry) {
      /* Reference from file scope to a pointer whose prefix indicates
         function-scope memory region kind.  For string entries this can
         happen even when the string is allocated in file-scope memory,
         because strings referenced from a function-scope memory region
         are numbered as honorary members of that region during IL output. */
      if (is_string_entry_kind(entry_kind)) {
        (void)fprintf(f_display, "**possible non-file-scope ptr** (%p)",
                      (a_void_ptr)entry_ptr);
      } else {
        (void)fprintf(f_display, "**NON FILE SCOPE PTR** (%p)",
                      (a_void_ptr)entry_ptr);
      }  /* if */

Maybe the positive branch of the inner if-statement shouldn't output anything, given the comment? Alternatively, drop the pointer value in that case.

(That comment could also use a "See assign_entry_number for details.")

[mborland]

An il_display.c change might be reasonable here. I'm seeing:

    if (displaying_file_scope_il && !is_file_scope_entry) {
      /* Reference from file scope to a pointer whose prefix indicates
         function-scope memory region kind.  For string entries this can
         happen even when the string is allocated in file-scope memory,
         because strings referenced from a function-scope memory region
         are numbered as honorary members of that region during IL output. */
      if (is_string_entry_kind(entry_kind)) {
        (void)fprintf(f_display, "**possible non-file-scope ptr** (%p)",
                      (a_void_ptr)entry_ptr);
      } else {
        (void)fprintf(f_display, "**NON FILE SCOPE PTR** (%p)",
                      (a_void_ptr)entry_ptr);
      }  /* if */

Maybe the positive branch of the inner if-statement shouldn't output anything, given the comment? Alternatively, drop the pointer value in that case.

(That comment could also use a "See assign_entry_number for details.")

I'll make the change. Thanks.

[daveedvdv-nvidia]

Looks good to me.

[mborland]

@daveedvdv-nvidia should this be merged now? I think you reviewed prior to the CI running green.

[daveedvdv-nvidia]

Yep. I'll do that next.