Is it possible to only include filename and line number for error-level logs?

The Trice tool shows the location information automatically using the information inside li.json. Transmitting the file names as runtime strings I would not recommend for performance reasons but you can try this in your code:

TRICE_ENTER_CRITICAL_SECTION triceS("err:%s", __FILE__); trice16("err::%d\n", __LINE__); TRICE_LEAVE_CRITICAL_SECTION

TRICE appears to lack traditional logging levels

What kind of syntax you expect? Please see the information about Trice format string tags inside the TriceUserManual. There is also a -logLevel CLI switch.

and doesn't offer a trice_with_fl(...) function that could conditionally include this information.

What do you mean with a missing trice_with_fl(...) function? Please give details.

TRICE seems to require direct calls where I want to record the location information, preventing abstraction through helper functions or macros. Is this correct? (Where defined rather than where invoked)

Would you please explain in more detail, what this questions means. The Trice tool log output contains per default location information and you can also have a binary or ASCII log file.

My current logging approach uses an example log_error("My Error %2X", value); which abstracts and automatically includes constants like MODULE and PROCESS, variables like TASK CONTEXT (from FreeRTOS), TIMESTAMP, and then appends the actual error message with formatted arguments with snprintfs. How can I achieve similar functionality in TRICE without sending the text for module and process as arguments?

I assume this is not possible out of the box, but we can think about it and maybe find a solution, by extension of the Trice functionality.

Right now you could do s.th. similar to this:

\\\ \file myModuleA.c

static void logModule(void){
    trice("myModuleA-" );
}

static void logProcess(void){
    switch(processID){
    case 1: trice("myProcessX-");  break;
    case 2: trice("myProcessY-");  break;
    default: trice("unknown_ProcessID-" ); break;
}

static void log_error(char* msg, int value ){
    TRICE_ENTER_CRITICAL_SECTION
    logModule();
    logProcess();
    trice("task_%d-timestamp_%d-", TASK_CONTEXT, TIMESTAMP);
    trice("%s", msg);
    trice(" %2X\n", value);
    TRICE_LEAVE_CRITICAL_SECTION
}

void aFunction(void){
   ... 
   log_error("My Error", 42);
   ...
}

This has the advantage having just one call, but the disadvantages compiling the error message into the target, the need to transmit it byte by byte and to have a somehow fixed formatting of the error value. Also the location information, file and line, are pointing into function log_error.

An alternative:

\\\ \file myModuleA.c

static void logModule(void){
    trice("myModuleA-" );
}

static void logProcess(void){
    switch(processID){
    case 1: trice("myProcessX-");  break;
    case 2: trice("myProcessY-");  break;
    default: trice("unknown_ProcessID-" ); break;
}

static void log_context(void){
    logModule();
    logProcess();
    trice("task_%d-timestamp_%d-", TASK_CONTEXT, TIMESTAMP);
}

void aFunction(void){
   ... 
   TRICE_ENTER_CRITICAL_SECTION log_context(); trice("My Error %2X\n", 42); TRICE_LEAVE_CRITICAL_SECTION
   ...
}

Disadvantage is to have 2 calls for each error and the need for protection, but the messages are not going into the targets, trasmitted as IDs and the location information is correct. Also free formatting of the error value(s).

If we create a Trice macro triceWithContext(...), then this macro could implicitely call the function log_context. (As very first idea)

For consistent log formatting, what's the best approach when I have predefined MODULE and PROCESS strings? Using something like Trice(" My Error string %c", myArg); for every thing we log seems error-prone if I need to manually type if every time. For us, logs aren't just human readable where a typo doesn't matter.

That's the same theme. We need to find a kind of generic/universal syntax.

One more info: If several trices build a single log line, thats when only the last one ends with "\n", the target timestamp, if any, and the location information are taken from the first Trice macro.

* Lack of string concatenation support to define constant string easily

Yes, the Trice preprocessing is very limited - I am not able to do it better. Maybe some additional tool could be written to construct combined strings for Trice macros like this:

Do not write:

 trice("<XMODULE><YPROCESS><ZLEVEL><ETC> My Error string %c", myArg);

Just write:

 triceCTX("My Error string %c", myArg);

run context insert ==>

 trice("myModuleX-myProcessY-myZLevel-myETCMy Error string %c", myArg); 

run trice insert ==>

 trice(iD(1234), "myModuleX-myProcessY-myZLevel-myETCMy Error string %c", myArg); 

run trice clean ==>

 trice("myModuleX-myProcessY-myZLevel-myETCMy Error string %c", myArg); 

run context clean ==>

 triceCTX(iD(1234), "My Error string %c", myArg); 

run trice clean ==>

 triceCTX("My Error string %c", myArg); 

* Potential line fragmentation with preemption in FreeRTOS

TRICE_ENTER_CRITICAL_SECTIONand TRICE_LEAVE_CRITICAL_SECTIONexist.

* Redundancy of work every time I want to log anything, few options to abstract

I understand your needs and i think this is an important extension. It would be great, if you put down s.th. like a specification draft, so that we can discuss the details.

Would using TRICE's deferral mechanism help with the multiple-call approach to prevent line fragmentation in preemptive environments?

I do not think so, deferred and direct mode are part of the Trice runtime and orthogonal to the compile-time Trice tool handling.

How should I handle argument size matching in TRICE? If I need to replace something like log_error("My Error %24s,%2X") without using excessive buffer space, what's the recommended approach? That I just need to know what I'm doing at all times and not miss anything?

Within one Trice macro you can only have a single %s dynamic string like trice("Today is %10s\n", weekday);, not mixed with other values OR 0-12 values, all having the same bitwidth, default is 32. So with default settings trice is the same as trice32.
To replace log_error("My Error %24s,%2X"); you need to write TRICE_ENTER_CRITICAL_SECTION trice("My Error %24s," msg); trice("%2X", value); TRICE_LEAVE_CRITICAL_SECTION.

Finally: Sorry for the delayed answer, but I realized your post only today.

UPDATE: Some thoughts on implementing structured logging as binary logging with Trice

• compile time information: file, line, function, compiler, module, build time, firmware version, machine, user, location, host OS, log level, fmt, ...

• runtime information: uptime, timestamp, hw serial, task ID, stack depth, event count, core ID, position, ...

• User may write trice("wrn:MyValueA %d, myValueB %d\n", 11, 22);

• A trice insert -fullContext command would change that line into trice(iD(123), "wrn:<file:main.c><line:123><function:initABC><taskID:%d>MyValueA %d, myValue %d", getTaskID(), 11, 22); for example according to a configuration file.

• A trice insert -runtimeContext command would only add the runtime information trice(iD(124), "wrn:<taskID:%d>MyValueA %d, myValue %d", getTaskID(), 11, 22);

• The compile time information goes only into a compileTimeContext.json file, similar to file and line already now inside li.json and is displayed by the Trice tool according to the configuration.

• So a triceConfiguration.json file could get specified, where the user selects, which compile time and runtime information should be added, how to obtain it and also how this information should look like.

• Example: TRice("MSG:day is %d\n", Val); -> with trice insert -fullContext this gets:

  trice(200), "context: [hw=%x] [core=%x] [log level=MSG] [fn=main] [build time=2025-06-10_12:34:56] [weekday is %d]\n", getHwSerial(), getCoreID(), Val);

• Runtime generated strings need their own separate Trice. triceS("name=%20s\n", sVal); type Trice logs can get compile time context added by format string extension, but not get runtime context added. Also including runtime generated strings as runtime context in to "normal" Trice logs is not possible within the same Trice statement. How to handle that in a clean way?

• Example: TRiceS("MSG:weekday is %10s\n", sVal); -> with trice insert -fullContext this gets:

  trice(201), "runtimeContext: [hw=%x] [core=%x]", getHwSerial(), getCoreID()); TRiceS(iD(203), "compileTimeContext:[log level=MSG] [fn=main] [build time=2025-06-10_12:34:56] [weekday is %10s]\n", sVal);