Add format checking to printf-type APIs #8601
Conversation
Correct types passed to debug() calls - as part of this, block size and erase size are changed to be 32-bit - using 64-bit variables for these will cause unnecessary code bloat. Many uses of bd_size_t in the BlockDevice API should really be uint32_t or size_t, but that would be a bigger API change.
|
/morph build |
Build : SUCCESSBuild number : 3525 Triggering tests/morph test |
Exporter Build : SUCCESSBuild number : 3142 |
|
I did try adjusting those BlockDevice things to Not sure it'd make a massive difference anyway - I fixed what seemed to be the most wasteful, dividing by a 64-bit variable block size when it's a constant 512, but even that only seemed to shave a few bytes off. (Although maybe a bit more trimming elsewhere would eventually allow a 64/64 divide routine to be omitted from the library). |
Test : FAILUREBuild number : 3310 |
|
/morph test |
Test : SUCCESSBuild number : 3313 |
Description
Toolchains are capable of performing format-checking on user-defined
printfvariants. Add the necessary directives to do this for some commonly-used mbed OS APIs:errormbed_error_printfdebug,debug_ifStream::printf,Stream::scanf,RawSerial::printfSome clean-ups of warnings generated by this change.
SDBlockDevicechanges are most wide-ranging - it's using 64-bit arithmetic excessively due tobd_size_tbeing used for all sorts of things that must be much smaller (being limited tosize_tby being in-memory, or just block sizes).Pull request type