Ville Juven

One option that comes to mind would be to add another template class like px4::atomic_from_isr which would be safe to use from interrupts. This approach is nothing new. After a...

Another option, which I already implemented, is to make two definitions for the critical sections, e.g. px4_critical.c / px4_usr_critical.c 1. Kernel uses the other to mask/unmask interrupts 2. User uses...

A third option that comes to mind would be to add a flag that enables the irq locks. In driver that needs IRQ locks: ``` #define PX4_ATOMIC_NEED_IRQ_LOCK #include ``` .......

> Would a quick call be easier? Are you on PX4 slack? https://px4.slack.com/ > > We also have a public dev call every Wednesday at 11 eastern (~ 10 minutes...

Converted to draft so this does not get accidentally merged

There seem to be link time errors now, I'm not sure why this has not happened before: /opt/gcc/bin/../lib/gcc/arm-none-eabi/9.3.1/../../../../arm-none-eabi/bin/ld: src/drivers/optical_flow/paa3905/libdrivers__optical_flow__paa3905.a(PAA3905.cpp.obj): in function `px4::atomic::store(unsigned long long)': /__w/PX4-Autopilot/PX4-Autopilot/build/airmind_mindpx-v2_default/../../platforms/common/include/px4_platform_common/atomic.h:91: undefined reference to `__atomic_store_8'

Hmm, it really does look like the missing symbols are nowhere to be found, .e.g. __atomic_store_8 is truly missing. I had assumed the GCC toolchain would implement these even though...

@dagar was there still something you wish for me to do regarding this ? I have tried to keep up with new drivers but have not noticed any that would...

Seems like my local version of nuttx is a bit newer, it contains this patch which implements getprogname() https://github.com/apache/incubator-nuttx/commit/ed1e4ddfa7d1dba8326421b27d16664feafa661e What is the process of incorporating patches from nuttx upstream ?...

> Should we wait until the upgrade? #20190 Thanks, sounds good!