Original Redmine Comment
Author Name: Dan Petrisko (@dpetrisko)
Original Date: 2019-05-16T00:48:57Z

$ verilator --lint-only test.v

%Warning-WIDTHCONCAT: test.v:8:

Unsized numbers/parameters not allowed in concatenations.

%Warning-WIDTHCONCAT: Use "/* verilator lint_off WIDTHCONCAT */" and lint_on around source to disable this message.

%Warning-WIDTHCONCAT: test.v:9: Unsized numbers/parameters not allowed in concatenations.

%Warning-WIDTHCONCAT: test.v:10: Unsized numbers/parameters not allowed in concatenations.

%Warning-WIDTHCONCAT: test.v:11: Unsized numbers/parameters not allowed in concatenations.

%Warning-WIDTHCONCAT: test.v:12: Unsized numbers/parameters not allowed in concatenations.

%Warning-WIDTHCONCAT: test.v:8: Unsized numbers/parameters not allowed in replications.

%Error: Exiting due to 6 warning(s) %Error: Command Failed /usr/bin/verilator_bin --lint-only test.v

The fix is to explictly size the enum value. But it seems like the typedef should take care of this

typedef enum logic[2:0] {P=3'd0, W, E, N, S} Dirs;

Original Redmine Comment
Author Name: Dan Petrisko (@dpetrisko)
Original Date: 2019-05-16T00:50:11Z

Verilator 4.015 devel rev UNKNOWN_REV

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2019-05-16T01:44:18Z

I suspect you added the LEN casts to try to work around this, but note they are unnecessary to show the problem, which can simply be seen with

typedef enum logic [2:0] {P=0};
localparam [5:0] SEQ = {P, P};

What is happening is the assignment of P to 0 is seen as unsized because the zero is unsized. Indeed if you use P=3'b0 it works. This is an unfortunate side effect of trying not to give WIDTH warnings when a user does "enum { P=0 }" then uses P in a context that is looking for only a few bit wide number (versus 32 bits wide).

Requires some experiments as how to fix this without adding those unfortunate WIDTH warnings.

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2019-07-06T20:27:54Z

Fixed in git towards 4.018.

I tried a few alternatives to preserve the behavior I thought was best for WIDTH warnings but ended up not finding anything that seemed consistent and didn't break this case (which seemed worse).

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2019-08-29T23:14:39Z

In 4.018.