Asynchronous reset width check in SDF appears to specify wrong edge

[alexroach-rl]

Hi All,

I came across a potential issue with OpenRoad's SDF generation where it was specifying the wrong edge for the width check of an active-low asynchronous reset pin.
IE The SDF contained the following:
(WIDTH (posedge RN) (...))

As I understand it, this is specifying the positive edge of RN as the reference event for the width check. This means that it would be checking the width of a high pulse on RN instead of the width of a low pulse, which is what you would normally care about for an active-low reset signal.

The lib file does not have a min_pulse_width_low constraint directly on the RN pin. Instead, the RN pin has a min_pulse_width timing section with a fall_constraint lookup table.
IE:

timing () {
  related_pin : "RN";
  timing_type : min_pulse_width;
  fall_constraint (template) {
    .
    .
  }  
}

[maliberty]

This should be opened as an issue against https://github.com/parallaxsw/OpenSTA with a test case.