intent/io-map-coverage

Remedy ¶

Declare the nets the map does not cover, or accept the coverage and read the other IO-map results as being about the covered fraction alone. There is nothing to fix in the design here.

What it checks ¶

Which nets in the design the IO map declares, and which it says nothing about. One verdict per net.

Nothing here is a defect. A net the map does not name is a question nobody asked, not a fault in the board. The rule exists so that number is visible rather than absent.

Why coverage is the number that matters here ¶

A design with sixteen hundred nets and a map declaring two hundred of them has two hundred checked and fourteen hundred unexamined, which is not the same as clean. Report only the two hundred and the result is green about a fraction of the board, with the size of that fraction nowhere in the output.

This is the other half of a pair, and the two get confused constantly because they look alike and are opposite defects:

	usually means	reported by
the map declares a net the netlist does not have	a real disconnection	intent/io-map-net-absent, as a fail
the netlist has a net the map does not declare	an incomplete map	this rule, as not-considered

Rails and grounds are counted, and said so ¶

A pin map does not usually name GND or a supply rail, and on a real board those are a large share of the nets. They are still counted in the denominator, and their verdict says which they are.

Carving them out would be the tool deciding which absences are acceptable, and that is the judgment that lets a real gap hide. A map that forgot an entire peripheral bank would read as well-covered if the arithmetic quietly excused a third of the board. Naming them in the reason gives a reader the discrimination without the tool making the call.

Reading the result ¶

The rule heading carries the whole tally, so the coverage number is the heading and not something to count by hand:

intent/io-map-coverage  47 pass, 1553 not-considered

The per-net rows are the addressable form of the same thing, so a specific net can be asked about rather than only the total. In the terminal the row list is capped, with the elided count stated; the heading is always complete.

Declaring it ¶

Nothing extra. It reads the same io_map section the other three rules read, and it is compiled whenever a map is declared.

Fixing a finding ¶

There is nothing to fix in the design. Either declare the nets the map does not cover, or accept the coverage and read the other IO-map results as being about the covered fraction alone.