Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2012-03-23T01:29:57Z

Obviously Verilator currently expects a format string.

Original Redmine Comment
Author Name: Alex Solomatnikov
Original Date: 2012-03-23T01:32:55Z

Another example:

$display($realtime, "ps Warning: Inclock_Period Violation");

Original Redmine Comment
Author Name: Alex Solomatnikov
Original Date: 2012-04-24T19:21:09Z

More $display() cases that don't work with verilator:

$display("Error!  lpm_representation value must be \"SIGNED\" or \"UNSIGNED\".", $time);

verilator output:

%Error: ...: Extra arguments for $display format

Another case:

$display("Time: %0t", $time, "   Warning : Detected transition on SCANCLK during quiet time. PLL may not function correctly."); 

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2015-10-21T01:15:40Z

See also #�.

Patches are welcome; at one point the code didn't discriminate between a user-string "..." and a number which just happened to print as a string, so it couldn't know if $display("foo") should print as a number or string. Some of that is now better, but it's likely still a problem in some places.

Also it's better to make a large format string than to break it up, because Verilator knows how to constant propagate into displays, so ideally a large $display will become in the output just a constant string. This doesn't work across multiple displays.

Original Redmine Comment
Author Name: Jamey Hicks
Original Date: 2015-10-21T13:40:01Z

I've started working on this. I see what you mean about constant propagation into display statements.

One question I have is how to tell if one of the arguments (other than the first) is a string.

I have this as input:

 $write("%h", a, ">");

The dtype for the node for the ">" dumps:

BASICDTYPE 0x2687090 <e17679> {g3222} @dt=this@(G/w8)  logic [GENERIC] kwd=logic range=[7:0]

Is there some clue that this was a string?

Original Redmine Comment
Author Name: Jamey Hicks
Original Date: 2015-10-21T13:41:20Z

My current plan is to extend the format string in LinkResolveVisitor with appropriate format % operators for the extra nodes.

Original Redmine Comment
Author Name: Jamey Hicks
Original Date: 2015-10-21T14:19:38Z

Found V3Number::isFromString(), answering my own question.

Original Redmine Comment
Author Name: Jamey Hicks
Original Date: 2015-10-21T16:48:03Z

Attached is a work in progress to address this issue. I'll remove all the chatter from the final patch.

As I append strings found in the argument list to the format string, I need to remove those strings from the argument list. Is argp->unlinkFrBack() the right thing to call?

I'll work on some tests and then clean it up.

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2015-10-23T00:30:20Z

Generally very good. Just a few nits.

+++ b/src/verilog.y
-       |       yD_FCLOSE '(' idClassSel ')'            { $$ = new AstFClose($1, $3); }
+       |       yD_FCLOSE '(' expr ')'                  { $$ = new AstFClose($1, $3); }
         |       yD_FFLUSH parenE                        { $1->v3error("Unsupported: $fflush of all handles does not map to C++."); }
-       |       yD_FFLUSH '(' idClassSel ')'            { $$ = new AstFFlush($1, $3); }
+       |       yD_FFLUSH '(' expr ')'                  { $$ = new AstFFlush($1, $3); }

Assume some tests will show the need for this, and that the change works OK.  Please provide as a separate patch.


+++ b/src/V3LinkResolve.cpp

-    void expectFormat(AstNode* nodep, const string& format, AstNode* argp, bool isScan) {
+    void expectFormat(AstNode* nodep, const string& format, AstNode* argp, bool isScan, string *newFormatp = 0) {

Please pass newFormat as a reference (newFormatr - r means reference).  Better, it can just be the return value.

+               int isFromString = 0;

bool


+                   int inpercent = 0;

bool inpercent = false;  (likewise "= true" below)

+                   newFormat.append(", %h");

newFormat += ", %h";

Original Redmine Comment
Author Name: Jamey Hicks
Original Date: 2015-10-26T15:17:17Z

Thanks for the feedback. Attached is the updated patch.

Original Redmine Comment
Author Name: Jamey Hicks
Original Date: 2015-10-26T15:42:18Z

One more update which supersedes the last patch.

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2015-10-28T00:59:33Z

Great, pushed to git towards 3.878.

(Made one trivial change to add a dataByte() accessor to V3Number, which seemed cleaner.)

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2015-11-01T13:22:26Z

In 3.878.

Original Redmine Comment
Author Name: John Demme (@teqdruid)
Original Date: 2016-12-22T21:42:57Z

Hi Guys-

Sorry, but from reading this exchange it's difficult for me to tell what's been fixed. Is the following now supported?
$display ($time, "ps Error: STRATIX or STRATIXGX in non DPA mode does not support the specified deserialization factor!");

I ask since similar things in altera_mf.v don't seem to work in the latest master. (This appears to be the last barrier to altera_mf.v parsing!)

~John

Original Redmine Comment
Author Name: Wilson Snyder (@wsnyder)
Original Date: 2017-01-10T00:19:56Z

The $display($time) is fixed in git towards 3.891.

Original Redmine Comment
Author Name: John Demme (@teqdruid)
Original Date: 2017-01-10T01:35:00Z

Excellent! Thanks