CONFIG_PROC_(LOG)_HAVE_CHRONO misleading

[beartronix]

SystemCore/Log.cpp

Line 33 in 5d4128c

#define CONFIG_PROC_LOG_HAVE_CHRONO 1

was previously CONFIG_PROC_HAVE_CHRONO and in Processing.h (better in my opinion because other Processes also (can) reference this define

[beartronix]

related: nowMs() and millis() are ambiguous:

SystemCore/Pipe.h

Line 60 in 5d4128c

#define nowMs() ((uint32_t)std::chrono::duration_cast<std::chrono::milliseconds>(std::chrono::system_clock::now().time_since_epoch()).count())

SystemCore/TcpTransfering.cpp

Line 754 in 5d4128c

uint32_t TcpTransfering::millis()

question again: why implement millis() everywhere instead of a central location? (or better: don't implement and leave it to the application code)

[Naegolus]

Thanks for the input! I will try to work on this asap.

Response to your question (no answer but more brainstorming)

I do not follow the DRY principle to 100%. It's more like 90%. A central location means that multiple units like A, B and C would be related to another unit called D for example. The strength of the coupling decides if code duplication is OK for me. If the coupling is low, then code duplication is allowed, otherwise it's a clear dependency of multiple units to one topic.
In general I would like to avoid a "Core Library". On the other hand there is one central unit of course: Processing.cpp

Another question is: Should the base Processing() provide minimal time-related functions?

One argument for yes could be: A process is a function that needs time for creating its output. For me a function has the signature y = f(x) whereas the signature of a process would be y = p(x, t). In this sense, time is tightly coupled to the concept of a process.
But time is considered by the function success() already. At the moment a process has an abstract state for mapping time: Not finished (Pending) and finished (Positive or not).

Important structure info

We have multiple system layers

1. SystemCore
   1.1 Processing.cpp
   1.2 Other cpp files
2. Concrete process libraries
3. Application / project specific processes

The core is divided into two layers. The file Processing.cpp must be compile-able for all systems. Simple and complex, with standard libraries and without. I have to think about it, how time should be integrated in the sources. At compile time or runtime.

Current situtation

TcpTransfering::millis() is intentionally a private function, which should not be used by external user code. But for some targets this is a problem which must be resolved, because the code still must be compiled.

Same strategy with CONFIG_PROC_LOG_HAVE_CHRONO. This is a private define localized in Log.cpp. User code should not reference it.

Pipe.h is one possible way for Inter-Process-Communication. This file has a public connection to time. At the moment I am wondering if this is a good place for a function / define called nowMs() :|

Result

This issue is important but will take some time.