I also see the filter system in canlog could be very useful as a general filtering system. I suggest to move that into the central CAN class, and allow such software filters to be attached to listeners, callbacks, and loggers.
By refactoring this, I think we can get some very useful functionality.
@Michael, one question, for you: What is the purpose of RegisterCallback (and the callbacks vs listeners, in general). Listeners get their frames delivered via a queue (so must have a task running). Callbacks are simple function callbacks running synchronously with the CAN task. From what I can see, the only user of callbacks is vehicle_twizy where CanResponder is set as the callback on CAN1. Why is that necessary, given that the standard vehicle object has IncomingFrameCan1() to receive frames. Is this required for performance (running synchronously with the CAN receive task), or some other reason why it can’t run from the vehicle task on queue receive of the frame? In other words, why can’t we just call CanResponder() from within IncomingFrameCan1()?
Regards, Mark.
On 11 Jun 2019, at 8:49 PM, Mark Webb-Johnson <mark@webb-johnson.net> wrote:
It has become a little duplicated:
struct CAN_frame_t{canbus* origin; // Origin of the frameCAN_FIR_t FIR; // Frame information recorduint32_t MsgID; // Message IDunion{uint8_t u8[8]; // Payload byte accessuint32_t u32[2]; // Payload u32 access (Att: little endian!)uint64_t u64; // Payload u64 access (Att: little endian!)} data;
esp_err_t Write(canbus* bus=NULL, TickType_t maxqueuewait=0); // bus: NULL=origin};
typedef struct
{
uint32_t timestamp;
canbus* bus;
CAN_LogEntry_t type;
union
{
CAN_frame_t frame;
CAN_status_t status;
char* text;
};
} CAN_LogMsg_t;
typedef struct{CAN_frame_t last;uint32_t rxcount;struct __attribute__((__packed__)){struct {uint8_t Ignore:1; // 0x01uint8_t Changed:1; // 0x02uint8_t Discovered:1; // 0x04uint8_t :1; // 0x08uint8_t :1; // 0x10uint8_t :1; // 0x20uint8_t :1; // 0x40uint8_t :1; // 0x80} b;uint8_t dc; // Data bytes changeduint8_t dd; // Data bytes discovereduint8_t spare;} attr;} re_record_t;
CAN_frame_t has origin for the can bus the message arrived on, while CAN_LogMsg_t has frame.origin and bus (presumably because bus is needed for status and text messages).
Then we have candump and canlog virtual interfaces (canlog supports ’trace’ and ‘crtd’, while candump supports ‘crtd’ and ‘pap’).
The users of these higher level structures are RETOOLS and CANLOG. I suggest the following changes:
- Have a single format conversion virtual class, and then initial implementations for pcap, crtd and trace. That can virtualise the conversion of a CAN message to be logged/transmitted/received to/from the external format. This can include a factory for registration of types (by textual name), as well as support functions for command registration (including whether they have support for reading, writing, or both). This just provides format conversion and does not provide any file handling or transmission functions. Canlog and Candump get merged to form this.
- Perhaps base this on CAN_LogMsg_t, but I suggest a more generic name (perhaps CAN_message_t, to differentiate that this is a can message rather than a simple frame). It is CAN_frame_t + additional support for stats and general text messages.
- Tidy up the CAN_LogMsg_t bus vs frame.origin issue.
Simply put, separate the formatting into a standalone virtual implementations, then leave retools and canlog to handle the actual RE stuff or file logging, or whatever else is required with the data that gets converted.
Does that make sense?
Regards, Mark.
On 11 Jun 2019, at 3:01 PM, Michael Balzer <dexter@expeedo.de> wrote:
Mark,
when implementing the can log framework I introduced the CAN_LogMsg_t specifically to capture the original timestamp of the frame, so it wouldn't get lost if
the logging task gets behind, and to transport other log data types than just frames to the loggers.
I would also expect the RE tools stream to produce event info & statistics entries like can log does, as that is very helpful in analysing. Maybe rebasing
candump on can log would be better?
Regards,
Michael
Am 11.06.19 um 08:37 schrieb Mark Webb-Johnson:
This really should be unified, and probably not too hard to do. Crazy to have two so similar frameworks. I think we can just have one virtual class for converting CAN frames to/from other formats, and then implementations for CRTD, PCAP, etc.
I do remember that last time I looked at this, canlog used CAN_LogMsg_t, and candump used CAN_frame_t (with CAN_LogMsg_t having some other fields for logging statistics, etc), and that made it non-trivial. Perhaps canlog should be built on top of candump?
I will have a look at it.
Regards, Mark.
On 6 Jun 2019, at 4:06 PM, Michael Balzer <dexter@expeedo.de> wrote:_______________________________________________
Steve,
Mark quoted from the RE tools candump_crtd class, which has its own crtd formatter.
For the canlog framework I currently get the timestamp from esp_log_timestamp(). Should be easy to change that to gettimeofday().
Regards,
Michael
Am 05.06.19 um 18:39 schrieb Stephen Casner:
Mark,--
If by "unix julian time" you mean a timestamp like 1559751698 (the
time as I write this), than that is not what Greg and I have observed
in the CRTD logs. Here are the first lines in our files:
151766.574 CXX Info Type:crtd; Path:'/sd/gps.crtd'; Filter:1:100-100; Vehicle:TR2N;
383603.293 CXX Info Type:crtd; Path:'/sd/location.crtd'; Filter:1:100-100; Vehicle:TR1N;
Yet OVMS does have the current time from NTP:
OVMS# time
Time Zone:
UTC Time: 2019-06-05 16:35:45 UTC
Local Time: 2019-06-05 16:35:45 GMT
Provider: ntp
PROVIDER STRATUM UPDATE TIME
*ntp 1 57 Wed Jun 5 16:35:44 2019
Who calls candump_crtd::get() with the timeval that goes into the
sprintf() you mention?
-- Steve
On Wed, 5 Jun 2019, Mark Webb-Johnson wrote:
The CRTD timestamps should be unix julian time (not uptime). Produced by this:_______________________________________________
sprintf(m_buf,"%ld.%06ld %sR%s %0*X",
time->tv_sec, time->tv_usec,
busnumber,
(frame->FIR.B.FF == CAN_frame_std) ? "11":"29",
(frame->FIR.B.FF == CAN_frame_std) ? 3 : 8,
frame->MsgID);
Once module gets time from the vehicle (or gps), it should show the correct timestamps.
Regards, Mark.
OvmsDev mailing list
OvmsDev@lists.openvehicles.com
http://lists.openvehicles.com/mailman/listinfo/ovmsdev
Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
Fon 02333 / 833 5735 * Handy 0176 / 206 989 26
_______________________________________________
OvmsDev mailing list
OvmsDev@lists.openvehicles.com
http://lists.openvehicles.com/mailman/listinfo/ovmsdev
OvmsDev mailing list
OvmsDev@lists.openvehicles.com
http://lists.openvehicles.com/mailman/listinfo/ovmsdev
--
Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal
Fon 02333 / 833 5735 * Handy 0176 / 206 989 26
_______________________________________________
OvmsDev mailing list
OvmsDev@lists.openvehicles.com
http://lists.openvehicles.com/mailman/listinfo/ovmsdev
_______________________________________________ OvmsDev mailing list OvmsDev@lists.openvehicles.com http://lists.openvehicles.com/mailman/listinfo/ovmsdev
-- Michael Balzer * Helkenberger Weg 9 * D-58256 Ennepetal Fon 02333 / 833 5735 * Handy 0176 / 206 989 26