CanFestival

Hi Luis,
> My compiler (for TI DSP F2812) complaints about the variable declaration
> at line 124 of pdo.c:
>
> Message pdo = {*pwCobId, REQUEST, 0};
>
> Because it only admits this type of initialization when values are
> constant. I have replaced with:
>
> Message pdo;
> pdo.cob_id.w = *pwCobId;
> pdo.rtr = REQUEST;
> pdo.len = 0;
>
> A patch is attached
>   
Now in CVS.

Edouard.

Hello Luis.

Your patch is very impressive. Very good work !

I will test it as soon as possible.

I have some suggestions :
- Add a paragraph in the tex manual (something similar to your mail
should be sufficient)
- Add some EMCY scenario to the TestMasterSlave example, so that :
  + the EMCY feature could also benefit from this non-regression test
  + CanFestival users will have a real use case to copy/paste code from.
- Add your name in CONTRIBUTORS file
- Put your name in the emcy.c and emcy.h files header
- 1003h must have a special behavior in objdictedit. See gen_cfile.py,
line 320 for a similar example:

    if 0x1005 not in communicationlist:
        entry_infos = Node.GetEntryInfos(0x1005)
        texts["EntryName"] = entry_infos["name"]
        indexContents[0x1005] = """\n/* index 0x1005 :   %(EntryName)s */
                    UNS32 %(NodeName)s_obj1005 = 0x0;   /* 0 */
"""%texts

    As you can see this forces declaration of a fake 1005 index in the
generated c file, so that CO_Data struct can be filled by initializer. 
Just do the same for 1003.

- EMCY_MAX_ERRORS should be a configure parameter:
    This do _not_ work (only first array element is initialized) :

	/* error_data: structure s_errors */\
	{{\
		0,          /* errCode */\
		0,          /* errRegMask */\
		0           /* active */\
	},},\


    Configure should generate config.h with appropriate macros :

#define EMCY_MAX_ERRORS 8

#define REPEAT_EMCY_MAX_ERRORS_TIMES(repeat)\
repeat repeat repeat repeat repeat repeat repeat repeat

And then, in data.h:

#define ERROR_DATA_INITIALIZER \
	{\
		0,          /* errCode */\
		0,          /* errRegMask */\
		0           /* active */\
	},

#define CANOPEN_NODE_DATA_INITIALIZER(NODE_PREFIX) {\
	...
	\
	/* EMCY */\
	Error_free,                      /* error_state */\
	sizeof(NODE_PREFIX ## _obj1003) / sizeof(NODE_PREFIX ## _obj1003[0]),      /* error_history_size */\
	& NODE_PREFIX ## _highestSubIndex_obj1003,    /* error_number */\
	& NODE_PREFIX ## _obj1003[0],    /* error_first_element */\
	& NODE_PREFIX ## _obj1001,       /* error_register */\
	/* error_data: structure s_errors */\
	{\
          REPEAT_EMCY_MAX_ERRORS_TIMES(ERROR_DATA_INITIALIZER)\
	},\
	_post_emcy              /* post_emcy */\


I send you CVS write access login/password in a separate email.

Thanks for your devotion to the CanFestival project !

Regards

Edouard

canfestival-devel@... a écrit :
> As I already commented in this list, I wanted to implement EMCY
> messages, and finally I have had time to do it.
> It is a preliminar version that only manages standard errors (no
> manufacturer specific field is used in EMCY messages or additional info
> in 1003 regs) but it seems to work.
> - It implements the error state machine defined in DS-301
> - It manages 1001h and 1003h OD entries. (It is neccesary that 1003h exists)
> - It uses an array (d->error_data) to manage which errors are active. By
> now the size of this array is defined as EMCY_MAX_ERRORS=8 in emcy.h,
> but it might be defined in config.h)
> - It calls the function (d->post_emcy(UNS8 nodeID, UNS16 errCode, UNS8
> errReg)) when an EMCY message is received. So it has to be defined in
> user app and assigned in the same way as initialisation() or post_sync().
> - It provides an API for error management:
>     · EMCY_setError(CO_Data* d, UNS16 errCode, UNS8 errRegMask)
>     · EMCY_errorRecovered(CO_Data* d, UNS16 errCode)
>
> I have faced some issues:
> - I haven´t found the criteria for MSG_ERR or MSG_WAR codes, so I used
> some arbitrary ones. It has to be changed.
> - Index 1003, subindex 0, number of errors: can be written, but only
> with a '0'. If it is tried to be written with a different value it
> should lead to an abort message that says it is out of range. I supose
> this verification has to be checked in the object access, but I don´t
> know where exactly, so I haven't done it.
>
> I have mad a patch, but it does not include "emcy.h" or "emcy.c" that
> are also attached.
>
> Please, let me know your impressions or suggestions if you test it.
> Regards,
>
>   
> ------------------------------------------------------------------------
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2005.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> ------------------------------------------------------------------------
>
> _______________________________________________
> Canfestival-devel mailing list
> Canfestival-devel@...
> https://lists.sourceforge.net/lists/listinfo/canfestival-devel


-- 
LOLITech - TBI sarl
24, rue Pierre Evrat
88100 Saint-Dié des Vosges
tel: 03 29 52 95 67
fax: 03 29 58 93 16
edouard.tisserant@... <mailto:edouard.tisserant@...>
http://www.lolitech.fr 	<http://www.lolitech.fr>

Hi, Edouard
I applied your suggestions yesterday and the changes are already in the
CVS. But I still have some questions:
- In the TeX manual, I have modified the line that indicates that EMCY
messages are active, but I don't know where to write down some more
instructions.
- The codes in MSG_ERR and MSG_WAR macros are arbitrary because I don´t
know how to assign the codes.
- And the most important problem: as specified in DS-301 about writing
to 1003h:
"Writing a 0 to sub-index 0 deletes the entire error history (empties
the array). Values higher than
0 are not allowed to write. This have to lead to an abort message (error
code: 0609 0030h)."
If I'm not wrong, there is not a general way of checking the range of
values for OD entries, so the easiest solution I have found is a
particular check for 1003 sub 0 in objacces.c that verifies that data ==
0. It would be after read-only check at line 288 approx in objacces.c
It is not a smart solution at all, but it would work. What do you think?
Regards,
-- 
Luis Jimenez
<http://www.suinsa.com/>

Hello Luis
> I applied your suggestions yesterday and the changes are already in the
> CVS. But I still have some questions:
>   
Thanks for this contribution. I'm sure users will appreciate it !
> - In the TeX manual, I have modified the line that indicates that EMCY
> messages are active, but I don't know where to write down some more
> instructions.
>   
You are right. Manual do still not have an API section. Doxygen is here 
for that, but at that time very small part of code comments have been 
doxyfied and filled.
> - The codes in MSG_ERR and MSG_WAR macros are arbitrary because I don´t
> know how to assign the codes.
>   
See CanFestival-3/doc/code-debug.txt. Sorry this is still not in the 
user manual.
> - And the most important problem: as specified in DS-301 about writing
> to 1003h:
> "Writing a 0 to sub-index 0 deletes the entire error history (empties
> the array). Values higher than
> 0 are not allowed to write. This have to lead to an abort message (error
> code: 0609 0030h)."
> If I'm not wrong, there is not a general way of checking the range of
> values for OD entries, so the easiest solution I have found is a
> particular check for 1003 sub 0 in objacces.c that verifies that data ==
> 0. It would be after read-only check at line 288 approx in objacces.c
> It is not a smart solution at all, but it would work. What do you think?
>   
There is a better solution. Use the _valueRangeTest stuff.

In the object dictionary editor :
- Add->UserType->choose UNSIGNED8, max 0, min 0

in generated code, it will correspond to:

#define valueRange_1 0xA0 /* Type UNS8, 0 < value < 0 */
UNS32 TestMaster_valueRangeTest (UNS8 typeValue, void * value)
{
  switch (typeValue) {
    case valueRange_1:
      if (*(UNS8*)Value < (UNS8)0) return OD_VALUE_TOO_LOW;
      if (*(UNS8*)Value > (UNS8)0) return OD_VALUE_TOO_HIGH;
      break;
  }
  return 0;
}

As you can see in the editor, you can only use this type in manufacturer 
specific section of the OD. Nevertheless, we can tweak gen_cfile.py a 
bit, so that :
- There is alway a valueRange_EMCY defined to a number inferior to 0xA0, 
so that it do not conflict with real user types.
- _valueRangeTest always have a valueRange_EMCY case, checking for ==0
- 1003h always use this type for first subindex, and first subindex is 
initialized to 0

We are implementing this with Laurent, it should be in CVS in one hour. 
Please test it.

Regards.

Edouard

-- 
LOLITech - TBI sarl
24, rue Pierre Evrat
88100 Saint-Dié des Vosges
tel: 03 29 52 95 67
fax: 03 29 58 93 16
edouard.tisserant@... <mailto:edouard.tisserant@...>
http://www.lolitech.fr 	<http://www.lolitech.fr>

Hello Michal.
> we have found, that recent CVS commit "Add win32 target (--can=win32) to 
> compile with msys and mingw32" made configure not running under dash shell. 
> The reason is that configure script contains some non-POSIX extensions and 
> the first line was changed #!/bin/bash -> #!/bin/sh. Since dash is used 
> as /bin/sh on ubuntu, I think, this is a big problem.
>   
You are right. I imagine is was a similar problem on debian ?

We reverted to #!/bin/bash, and we will update the manual to tell mingw 
users to install bash.
> I put some corrections to the attached patch, but I sill don't know correct 
> substitution for ${!i}. It seems to me, that plain POSIX doesn't support 
> variable indirection - neither ${!i} nor ${${i}} (see 
> http://www.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html#tag_02_06_02)
>   
If we revert to #!/bin/bash, do you consider you changes still worth to 
be in CVS? What do they bring to end user ? Which platform do not have 
bash ?

Thanks for your feedback,

Edouard
-- 
LOLITech - TBI sarl
24, rue Pierre Evrat
88100 Saint-Dié des Vosges
tel: 03 29 52 95 67
fax: 03 29 58 93 16
edouard.tisserant@... <mailto:edouard.tisserant@...>
http://www.lolitech.fr 	<http://www.lolitech.fr>

On Wednesday 19 of September 2007, canfestival-devel@... 
wrote:
> Hello Michal.
>
> > we have found, that recent CVS commit "Add win32 target (--can=win32) to
> > compile with msys and mingw32" made configure not running under dash
> > shell. The reason is that configure script contains some non-POSIX
> > extensions and the first line was changed #!/bin/bash -> #!/bin/sh. Since
> > dash is used as /bin/sh on ubuntu, I think, this is a big problem.
>
> You are right. I imagine is was a similar problem on debian ?

Hello,

It seems that Debian uses bash as /bin/sh, so no problems there.
>
> We reverted to #!/bin/bash, and we will update the manual to tell mingw
> users to install bash.
>
OK.
> > I put some corrections to the attached patch, but I sill don't know
> > correct substitution for ${!i}. It seems to me, that plain POSIX doesn't
> > support variable indirection - neither ${!i} nor ${${i}} (see
> > http://www.opengroup.org/onlinepubs/009695399/utilities/xcu_chap02.html#t
> >ag_02_06_02)
>
> If we revert to #!/bin/bash, do you consider you changes still worth to
> be in CVS? What do they bring to end user ? Which platform do not have
> bash ?

If you stay with bash, my changes are not important. I only think, that use of 
== instead of = is typo. According to bash manual, == is used to match 
against a pattern, but there is no pattern in the right operand.

Michal

Hello Michal.
> If you stay with bash, my changes are not important. I only think, that use of 
> == instead of = is typo. According to bash manual, == is used to match 
> against a pattern, but there is no pattern in the right operand.
Fixed in CVS.

Thanks for your feedback,

Edouard

Hi Tim,

I didn't address this problem yet.

-Leonid

canfestival-devel@... wrote:
> Hello Leonid,
>
> I'm preparing some tests at the moment. Are there changes regarding 
> the Ixxat read_sdo problem?
>
> Maybe I can make some more checks on the driver?
>
> Best regards
>
> Tim
> ------------------------------------------------------------------------
>
> -------------------------------------------------------------------------
> This SF.net email is sponsored by: Microsoft
> Defy all challenges. Microsoft(R) Visual Studio 2005.
> http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/
> ------------------------------------------------------------------------
>
> _______________________________________________
> Canfestival-devel mailing list
> Canfestival-devel@...
> https://lists.sourceforge.net/lists/listinfo/canfestival-devel
>   

-- 
Leonid Tochinski
Senior Software Architect
ChattenAssociates, Inc.
610-940-6040 x 118
ltochinski@...

Chatten Associates Head-Aimed Vision Systems ... Remote Vision for Robotics.

Dear Simone,
> In my application I have a node that cyclical transmits n to TPDO to regular
> intervals (125 ms). in the node there are not neither l'heartbeat neither
> the node guarding. I have the necessity of knowing if this node it is alive
> or not therefore I must "count" how many ms are pass from last message
> reception.I have read in cia 301 "... Independent of the trasmission type
> the RPDO event timer is used recognize the expiration of the RPDO", I
> suppose that if I do not receive that RPDO within the Time event, that RPDO
> is not valid, therefore all the variables mapped in that RPDO do not have to
> be considered valid.
> For me this is perfect, so I decide to implement the
> time event in the RPDO.
>   
OD entries (and subsequently mapped variables) do not have any "valid" 
bit. Nevertheless, RPDO do have such a "valid" bit : bit 31 of subindex 
00h of Receive PDO parameters (1400h - 15ffh).

Invalidating RPDO on timer expiration would imply CanFestival node to 
ignore data once timer as expired.

When timer expires it disables the PDO, and must also call an 
application "RPDO_Timeout_Error" callback. If callback is not filled by 
app, PDO will keep disabled until the app re-enable it with SetODEntry.
> I see that in canfestival this option is not implemented . Have you got any
> idea how to implement this option ?
>   
Read pdo.c in details. It should be clear to you at the end. If you need 
help, just ask.

Regards,

Edouard

-- 
LOLITech - TBI sarl
24, rue Pierre Evrat
88100 Saint-Dié des Vosges
tel: 03 29 52 95 67
fax: 03 29 58 93 16
edouard.tisserant@... <mailto:edouard.tisserant@...>
http://www.lolitech.fr 	<http://www.lolitech.fr>