Bug in ble_dfu_bonded.c?

[F Iannce]

I think I found a bug in ble_dfu_bonded.c.

In ble_dfu_buttonless_bootloader_start_prepare() there is a call to gscm_local_database_has_changed().  The problem is that this function is only present if PM_SERVICE_CHANGED_ENABLED is 1.  If this has been disabled, then you will get a linker error.

I removed the linker error by putting a #if(PM_SERVICE_CHANGED_ENABLED == 1) #endif around the gscm_local_database_has_changed() function call.

Is PM_SERVICE_CHANGED_ENABLED required for bonded DFU support?  I disabled it because I am running close to the flash limit even with optimizations...

[Vidar Berg]

I would not recommend disabling "service changed" support in the peer manager module unless you can replace it with something that has a smaller flash footprint. It's important that the bonded peripheral sends a service changed indication when the attribute table changes to ensure that the client performs new service discovery (E.g., when connecting to new app after DFU). Recommend reading the section about attribute caching in Bluetooth core specification for more details ( v.5.0, vol. 3, part G, section 2.5.2)

[F Iannce]

Ok.  I got some additional FLASH back from disabling the logging.  It'll be enough for now.  I'll re-enable the service changed feature.

A precompiler guard around this function call, however, should probably be there so that if someone does want to disable this feature they don't get compile errors...

Thanks,

-F