| Add Proposal | Add Analysis | Edit Class, Environment, or Release |
| Number | 423
|
| Category | errata
|
| Synopsis | 17.2.4.3 $sscanf and x/z
|
| State | open
|
| Class | errata-discuss
|
| Arrival-Date | Aug 19 2003
|
| Originator | sharp@cadence.com
|
| Release | 2001b
|
| Environment |
|
| Description |
In section 17.2.4.3 (bottom of page 292 of printed LRM), it says "If the format string, or the str argument to $sscanf contains unknown bits (x or z) then the system task shall return EOF." This is not consistent with the behavior of other Verilog system tasks and functions. When they get an x or z bit in an argument that is treated as a string or a real, they generally treat the bit as zero. Also, $sscanf is not a system task, it is a system function. It is particularly silly to call it a system task in a sentence which is talking about its return value. |
| Fix |
|
| Audit-Trail |
From: Shalom.Bresticker@motorola.com To: sharp@cadence.com Cc: etf-bugs@boyd.com Subject: Re: errata/423: 17.2.4.3 $sscanf and x/z Date: Wed, 20 Aug 2003 05:16:37 +0300 (IDT) Steven, > This is not consistent with the behavior of other Verilog > system tasks and functions. When they get an x or z bit > in an argument that is treated as a string or a real, they > generally treat the bit as zero. I agree about the case of a "real", but where is it written about the case of a "string" ? Thanks, Shalom -- Shalom Bresticker Shalom.Bresticker@motorola.com Design & Reuse Methodology Tel: +972 9 9522268 Motorola Semiconductor Israel, Ltd. Fax: +972 9 9522890 POB 2208, Herzlia 46120, ISRAEL Cell: +972 50 441478 From: Steven Sharp <sharp@cadence.com> To: sharp@cadence.com, Shalom.Bresticker@motorola.com Cc: etf-bugs@boyd.com Subject: Re: errata/423: 17.2.4.3 $sscanf and x/z Date: Wed, 20 Aug 2003 18:49:36 -0400 (EDT) >I agree about the case of a "real", >but where is it written about the case of a "string" ? It is not written anywhere, as with many obscure corner cases in the language. However, I just did looked at the list of system tasks and functions and tested the obvious ones in Verilog-XL. $display with a %s format will print the string as if x and z were zero bits. $fopen will treat x and z in the filename as zero bits. $test$plusargs will treat x and z in the argument string as zero bits. And I believe that when PLI requests a value in string format, x and z bits get converted to zero bits. This provides precedent for handling them this way. I wouldn't mind if the behavior for $sscanf were left unspecified, as it was for these other cases. However, if the behavior is specified, it should be consistent with this precedent. BTW, NC-Verilog does not get the same result as XL for all of these cases. Since the behavior is not actually specified, it is questionable whether these can be considered a bugs or not. If the behavior were to be specified, then the specification should match XL, and NC would need to be fixed. Steven Sharp sharp@cadence.com |
| Unformatted |
|
Hosted by Boyd Technology