| Number | 242
|
| Category | errata
|
| Synopsis | 23.9, 27.15: Is -f standard ?
|
| State | lrmdraft
|
| Class | errata-ptf
|
| Arrival-Date | Jan 01 2003
|
| Originator | Shalom.Bresticker@motorola.com
|
| Release | 2001b: 23.9, 27.15
|
| Environment |
|
| Description |
Is the -f command-line argument standardized by 1364-2001? The PLI routines in 23.9 and 27.15 seem to relate to it as something standard, or at the very least seem to say that if you use it, it must behave it in a certain way. Yet, there is no mention of this anywhere else in the LRM. Looking in my mail archives, I found that there were discussions about this, but I did not find a definite resolution to this. It seems to have ceased to be a subject for discussion at some point. (I don't have a PTF archive, and my VSG archive is only partial. My BTF archive is almost complete.) |
| Fix |
Section 23.9 acc_fetch_argv page 425 (second paragraph) Replace : The argument following a -f argument shall contain a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file. The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be a NULL. If one of the options is a -f then the next pointer shall behave the same as described above. With: "The vendor tool may provide a command line option to pass a file containing a set of options. In that case, the argument strings returned by acc_fetch_argv() shall contain the vendor option string name followed by a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file. The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be NULL. If one of the options is the vendor file option, then the next pointer shall behave the same as described above. Section 27.15 vpi_get_vlog_info page 682 last paragraph: Replace : The argument following a -f argument shall contain a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file. The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be a NULL. If one of the options is a -f then the next pointer shall behave the same as described above. With: "The vendor tool may provide a command line option to pass a file containing a set of options. In that case, the argument strings returned by vpi_get_vlog_info() shall contain the vendor option string name followed by a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file. The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be NULL. If one of the options is the vendor file option, then the next pointer shall behave the same as described above. |
| Audit-Trail |
Analyzed by etf@boyd.com on Thu Mar 27 13:54:54 2003 Detailed analysis: The PLI function acc_fetch_argv and VPI function vpi_get_vlog_info have the following description: " The format of the argv array is that each pointer in the array shall point to a NULL terminated character array which contains the string located on the tool's invocation command line. There shall be ‘argc’ entries in the argv array. The value in entry zero shall be the tool's name. The argument following a -f argument shall contain a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file. The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be a NULL. If one of the options is a -f then the next pointer shall behave the same as described above. " The issue is that the standard does not standardize on any command line argument, except for the + args but even there it is only the + which is part of the standard. The PLI section seems to indicate that the argument -f is standard and is followed by the name of a file containing a set of options. Proposal: remove the mention of a -f argument and state that the simulator tool may provide a command line option to provide a file containing a set of options. In that case the acc_fetch_argv and vpi_get_vlog_info will contain this option name followed by a pointer to null terminated array of pointers to characters. Resolution Replace : The argument following a -f argument shall contain a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file. With: "the simulator tool may provide a command line option to provide a file containing a set of options. In that case the acc_fetch_argv and vpi_get_vlog_info will provide the option string name followed by a pointer to null terminated array of pointers to characters. This new array shall contain the parsed contents of the file" Another alternative is to agree on a identifier marker such as <option_file> to replace the "-f" which would semantically identify the fact that the following array of strings have been extracted from an option file. This marker identifier will be output by acc_fetch_arg and vpi_get_vlog_info and will replace any command line option name chosen by a given simulator. From: Shalom Bresticker <Shalom.Bresticker@motorola.com> To: ptf-bugs@boyd.com Cc: Subject: Re: errata/242: PROPOSAL - 23.9, 27.15: Is -f standard ? Date: Sun, 30 Mar 2003 15:15:35 +0300 Resend: Date: Thu, 27 Mar 2003 14:14:45 -0800 From: Stephen Williams<steve@icarus.com> > "the simulator tool may provide a command line option to provide a > file containing a set of options. In that case the acc_fetch_argv and > vpi_get_vlog_info will provide the option string name followed by a > pointer to null terminated array of pointers to characters. This new > array shall contain the parsed contents of the file" Seems to be wrong because it legislates away portability. A programmer needs to know, when scanning the arguments, that the argument is a pointer to a sub-array instead of a char*. The prefix to this array pointer needs to be standardized, or standard argument scanners are theoretically impossible. Date: Fri, 28 Mar 2003 05:53:00 +0300 (IDT) From: Shalom.Bresticker@motorola.com What would be the problem in making that prefix customizable? In point of fact, -f is today not universally standard. Date: Thu, 27 Mar 2003 19:00:27 -0800 From: Stephen Williams<steve@icarus.com> > What would be the problem in making that prefix customizable? > In point of fact, -f is today not universally standard. The PLI code would need to be informed what that flag is so that it could properly traverse the argument list. If it assumes "-f" for one Verilog, and it works, then that same PLI program could crash when run by another compiler that uses a different prefix. Proposal #2 would not suffer this failing.: > "the simulator tool may provide a command line option to denote a file > containing a set of options. In that case the acc_fetch_argv and > vpi_get_vlog_info will provide the string name "<option_file>" > followed by a pointer to null terminated array of pointers to > characters. This new array shall contain the parsed contents of the > file" I believe this would work. From: Debi Dalio <debid@model.com> To: ptf-bugs@boyd.com Cc: Subject: [Fwd: Re: errata/242: PROPOSAL - 23.9, 27.15: Is -f standard ?] Date: Mon, 31 Mar 2003 09:49:20 -0800 >Category: errata >Confidential: no >Originator: Debi Dalio <debid@model.com> >Release: 2001b >Class: TBD >Description: This is a multi-part message in MIME format. --------------050603010605040407030207 Content-Type: text/plain; charset=us-ascii; format=flowed Content-Transfer-Encoding: 7bit Precedence: bulk Just to thrown my 2 cents in here regarding the handling of -f in acc_fetch_argv and vpi_get_vlog_info: ModelSim does not maintain the knowledge that a -f was used. It simply concatenates all simulator options into one char * array. Therefore, for acc_fetch_argv and vpi_get_vlog_info a simple char * array is returned. I don't see any particular reason why anyone would need to know that a -f was used; I would think it is mostly important just to know which command line options were specified, no matter where they came from. Keep it simple and and just have these routines return simple char * arrays. Debi -- Debi Dalio Staff Engineer office: 503-685-0856 Model Technology fax: 8105 SW Boeckman Rd email: debid@model.com Wilsonville, OR 97070-7777 USA web: www.model.com --------------050603010605040407030207 Content-Type: message/rfc822; name="Re: errata/242: PROPOSAL - 23.9, 27.15: Is -f standard ?" Content-Transfer-Encoding: 7bit Content-Disposition: inline; filename="Re: errata/242: PROPOSAL - 23.9, 27.15: Is -f standard ?" Received: from relay1.mentorg.com ([192.94.38.131]) by out.model.com with SMTP (Microsoft Exchange Internet Mail Service Version 5.5.2653.13) id HQ7BXVQC; Sun, 30 Mar 2003 04:20:04 -0800 Received: from gw.boyd.com ([64.171.19.218] helo=max.boyd.com) by relay1.mentorg.com with esmtp id 18zbnD-0007in-00 from owner-ptf@boyd.com for debid@model.com; Sun, 30 Mar 2003 04:20:07 -0800 Received: (from majordomo@localhost) by max.boyd.com (8.11.0/8.11.0) id h2UCK5W16398 for ptf-list; Sun, 30 Mar 2003 04:20:05 -0800 Received: (from mail@localhost) by max.boyd.com (8.11.0/8.11.0) id h2UCK3W16392 for ptf@boyd.com; Sun, 30 Mar 2003 04:20:03 -0800 Received: (from gnats@localhost) by max.boyd.com (8.11.0/8.11.0) id h2UCK2Y16382; Sun, 30 Mar 2003 04:20:02 -0800 X-Authentication-Warning: max.boyd.com: majordomo set sender to owner-ptf@boyd.com using -f Date: Sun, 30 Mar 2003 04:20:02 -0800 Message-Id: <200303301220.h2UCK2Y16382@max.boyd.com> To: ptf-bugs@boyd.com From: Shalom Bresticker <Shalom.Bresticker@motorola.com> Subject: Re: errata/242: PROPOSAL - 23.9, 27.15: Is -f standard ? Sender: owner-ptf@boyd.com Precedence: bulk The following reply was made to PR errata/242; it has been noted by GNATS. From: Shalom Bresticker <Shalom.Bresticker@motorola.com> To: ptf-bugs@boyd.com Cc: Subject: Re: errata/242: PROPOSAL - 23.9, 27.15: Is -f standard ? Date: Sun, 30 Mar 2003 15:15:35 +0300 Resend: Date: Thu, 27 Mar 2003 14:14:45 -0800 From: Stephen Williams<steve@icarus.com> > "the simulator tool may provide a command line option to provide a > file containing a set of options. In that case the acc_fetch_argv and > vpi_get_vlog_info will provide the option string name followed by a > pointer to null terminated array of pointers to characters. This new > array shall contain the parsed contents of the file" Seems to be wrong because it legislates away portability. A programmer needs to know, when scanning the arguments, that the argument is a pointer to a sub-array instead of a char*. The prefix to this array pointer needs to be standardized, or standard argument scanners are theoretically impossible. Date: Fri, 28 Mar 2003 05:53:00 +0300 (IDT) From: Shalom.Bresticker@motorola.com What would be the problem in making that prefix customizable? In point of fact, -f is today not universally standard. Date: Thu, 27 Mar 2003 19:00:27 -0800 From: Stephen Williams<steve@icarus.com> > What would be the problem in making that prefix customizable? > In point of fact, -f is today not universally standard. The PLI code would need to be informed what that flag is so that it could properly traverse the argument list. If it assumes "-f" for one Verilog, and it works, then that same PLI program could crash when run by another compiler that uses a different prefix. Proposal #2 would not suffer this failing.: > "the simulator tool may provide a command line option to denote a file > containing a set of options. In that case the acc_fetch_argv and > vpi_get_vlog_info will provide the string name "<option_file>" > followed by a pointer to null terminated array of pointers to > characters. This new array shall contain the parsed contents of the > file" I believe this would work. --------------050603010605040407030207-- ===================== previous proposals ===================== Proposal #1: remove the mention of a -f argument and state that the simulator tool may provide a command line option to provide a file containing a set of options. In that case the acc_fetch_argv and vpi_get_vlog_info will contain this option name followed by a pointer to null terminated array of pointers to characters. Resolution Section 23.9 acc_fetch_argv page 425 (second paragraph) Replace : The argument following a -f argument shall contain a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file.The value in entry zero shall contain the name of the file. With: "the vendor tool may provide a command line option to pass a file containing a set of options. In that case, the argument strings returned by acc_fetch_argv will contain the vendor option string name followed by a pointer to null terminated array of pointers to characters. This new array shall contain the parsed contents of the file. The value in entry zero shall contain the name of the file. Proposal #2: Agree on a string identifier which will be output by acc_fetch_arg and vpi_get_vlog_info and which will be substituting any vendor specific command line switch denoting an option file. Replace : The argument following a -f argument shall contain a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file. With: "the simulator tool may provide a command line option to denote a file containing a set of options. In that case the acc_fetch_argv and vpi_get_vlog_info will provide the string name "<option_file>" followed by a pointer to null terminated array of pointers to characters. This new array shall contain the parsed contents of the file" ------------------------------------------------------------ After debate on 4.7 we opted for solution 1 (vendor option string). options are not portable between tools so it is not an issue to provide back the vendor option name for file inclusion of options. We cannot flaten the options set into 1 set because of the way it is currently specified in the standard. Final resolution is: Section 23.9 acc_fetch_argv page 425 (second paragraph) Replace : The argument following a -f argument shall contain a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file.The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be a NULL. If one of the options is a -f then the next pointer shall behave the same as described above. With: "The vendor tool may provide a command line option to pass a file containing a set of options. In that case, the argument strings returned by acc_fetch_argv will contain the vendor option string name followed by a pointer to null terminated array of pointers to characters. This new array shall contain the parsed contents of the file. The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be a NULL. If one of the options is the vendor file option, then the next pointer shall behave the same as described above. Section 27.15 vpi_get_vlog_info page 682 last paragraph: Replace : The argument following a -f argument shall contain a pointer to a NULL terminated array of pointers to characters. This new array shall contain the parsed contents of the file.The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be a NULL. If one of the options is a -f then the next pointer shall behave the same as described above. With: "The vendor tool may provide a command line option to pass a file containing a set of options. In that case, the argument strings returned by vpi_get_vlog_info will contain the vendor option string name followed by a pointer to null terminated array of pointers to characters. This new array shall contain the parsed contents of the file. The value in entry zero shall contain the name of the file. The remaining entries shall contain pointers to NULL terminated character arrays containing the different options in the file. The last entry in this array shall be a NULL. If one of the options is the vendor file option, then the next pointer shall behave the same as described above. ------------------------------------------------------------ The proposal passed by the PTF on 6/16/2003 is in the fix field |
| Unformatted |
|
Hosted by Boyd Technology