Closed tommikas closed 8 years ago
@jupe How does this look? The mbed-trace-PR/label=linux,target=x86-linux-native build seems to be failing due to coverage again, but it actually seems unchanged looking at lcov->genhtml output. Does the coverage limit increase in ARMmbed/mbed-trace#25 already affect the whole repository?
@korjaa Fyi, this is what it looks like at the moment.
@jupe @tommikas Can this be handled differently and not by config.json ? Yotta is going away and this will soon be deprecated, we should move this logic to some header files.
The yotta functionality is just one alternative at the moment. It doesn't need to be used, and it can be easily removed whenever without affecting the other functionality. Happy to leave it out if it's considered unnecessary/unwanted, though.
Non-yotta builds are accommodated through 3 definitions, which are effectively synonymous to the yotta configuration options:
MBED_TRACE_MEM_INCLUDE
MBED_TRACE_MEM_ALLOC
MBED_TRACE_MEM_FREE
They make it possible to define whatever implementation is desired, including nsdynmemLIB. If nsdynmemLIB (for example) is the only alternative to stdlib that anyone using this library might want to use, I can put that behind a single flag or something, with stdlib remaining the default option.
@yogpan01 Did my previous comment address your concerns? I'm not sure if you saw it...
@korjaa Would this be sufficient for you? Even if you don't need to follow the master branch at the moment, might as well make it usable now rather than tinker with it later.
lgtm
+1 looks good, cannot think any improvements at the moment.
lgtm
Allows configuring the used implementation with a config.json entry of the form:
or by the usual means of passing in preprocessor directives in C builds using the definitions
If nothing is configured, defaults to stdlib.