search

armink / CmBacktrace

Advanced fault backtrace library for ARM Cortex-M series MCU | ARM Cortex-M 系列 MCU 错误追踪库
MIT License
1.58k stars 640 forks source link

Let firmware info be optional to save RAM #83

Open satur9nine opened 3 weeks ago

satur9nine commented 3 weeks ago

Many projects already have another mechanism for getting firmware info or want to save RAM and do not need firmware info of CmBacktrace.

satur9nine commented 6 days ago

@armink any thoughts on this?

armink commented 5 days ago

These version number information seems to be necessary, and optimization is not recommended

satur9nine commented 3 days ago

I agree that some system to track the version of product HW/SW version is necessary for most embedded systems but for example my company has been using CmBacktrace for many years but we already have our own version number mechanism and prefer not to have a duplicate system, thus it is not necessary for us and probably many other users. CmBacktrace works very well without version number tracking mechanism but your choice to take the PR or not.

armink commented 3 days ago

You can try to change CMB_USING_FIRMWARE_INFO to CMB_NOT_USING_FIRMWARE_INFO, so that everyone can upgrade without feeling. Then, you need to check whether the program works properly when this macro is turned on and off.

BenBE commented 2 days ago

Why not have a weak symbol cm_version_printinfo which resolves to a blank function if internal version info is disabled or a full function doing the output when enabled. If a user has its own mechanism they simply write their own function of that name without marking it as a weak symbol, thus replacing that function at link time.

Also, why have these version strings live inside of RAM buffers in the first place?

satur9nine commented 1 day ago

Why not have a weak symbol cm_version_printinfo which resolves to a blank function if internal version info is disabled or a full function doing the output when enabled. If a user has its own mechanism they simply write their own function of that name without marking it as a weak symbol, thus replacing that function at link time.

Also, why have these version strings live inside of RAM buffers in the first place?

Both good points, I can explore making these suggested changes.