Some ntdrivers-simplified have tautological comparisions

[sosy-lab]

Some ntdrivers-simplified have tautological comparisions

[mchalupa]

Are tautological comparisons a problem? (except that they flood the output when compiling with warnings)

[PhilippWendler]

Such warnings can show potential problems with the benchmarks (e.g., if there is some x >= 0 in the code and we change x to be unsigned because of overflows), but IMHO for old code like this here we can probably ignore them. At least I wouldn't know whether there is something to fix here, and the people that have created these benchmarks are no longer around. And the warnings here do not seem to indicate an issue with the validity of the benchmark.

[tautschnig]

Looking at ntdrivers/cdaudio_false-unreach-call.i.cil.c, which ought to be the more complete version of that benchmark, there is nothing to suggest that status == status even remotely makes sense.

        if (! irpSp->Parameters.UsageNotification.InPath) {
          if (deviceExtension->PagingPathCount == 1UL) {
            if (DeviceObject->Flags & 16384UL) {

            } else {
              DeviceObject->Flags |= 8192UL;
              setPagable = 1;
            }
          } else {

          }
        } else {

        }

Has become

        if (irpSp__Parameters__UsageNotification__InPath) {
          if (deviceExtension__PagingPathCount != 1) {
            goto _L;
          }
        } else {
          _L:
          if (status == status) {
            //DeviceObject__Flags |= 8192;
            setPagable = 1;
          }
        }

That is, somehow all the branching conditions have been inverted. The closest match of status == status would be DeviceObject->Flags & 16384UL. Maybe code in "Simple" just shouldn't be sent to GCC/Clang for checking as it's just too manipulated.