Refactor IWDG driver and add initialize function for prescaler with downcounter #1127
Conversation
salkinium
force-pushed
the
feature/init_iwdg_by_time
branch
from
8abf5f9
to
5a4a9bb
Compare
| out_index = idx; | ||
| } | ||
| }; | ||
| for (uint32_t idx{0}; const auto prescaler : prescalers) |
There was a problem hiding this comment.
Maybe
| for (uint32_t idx{0}; const auto prescaler : prescalers) | |
| for (const auto [idx, prescaler] : enumerate(prescalers)) |
?
There was a problem hiding this comment.
oh, std::ranges::enumerate_view exists, so we can alias this to modm::enumerate.
There was a problem hiding this comment.
👍, but that requires gcc 13. Mac OS hosted and AVR currently have 12. Let's upgrade the compilers for the next release and do it afterwards.
There was a problem hiding this comment.
We need to find a workaround for the Protothread macro issue to upgrade to GCC13…
There was a problem hiding this comment.
I don't think this is fixable. The whole construct relies on jumping into a statement expression with a case label. According to the gcc documentation this was never valid but is now properly diagnosed since gcc 13. It just happened to work often enough ...
The only acceptable solution I see is abandoning resumable functions.
There was a problem hiding this comment.
This doesn't seem to work due to an (for me) incomprehensible compiler error about unsigned int& lvalue vs unsigned int rvalue… help (see CI for issue)
There was a problem hiding this comment.
It's a limitation in modm::enumerate. The iterator types you have defined (linear_range, etc.) return r-values, here plain integers, not l-value references when you dereference them and the implementation can't deal with that. I'll push a fix tonight.
There was a problem hiding this comment.
It's fixed now. I've also pushed a fixup for a typo in the Nucleo G070 module.lb file.
salkinium
force-pushed
the
feature/init_iwdg_by_time
branch
3 times, most recently
from
3ec43f7
to
893a7f3
Compare
salkinium
force-pushed
the
feature/init_iwdg_by_time
branch
from
893a7f3
to
23036e3
Compare
Re-opening this since I prematurely merged the previous PR #1122 due to pushing the wrong branch (PEBCAK). Addresses the last review. Original description:
Ok, this escalated a little. I was just looking through the code and saw that the Iwdg class was not inlined, so I quickly refactored it and then noticed that the prescaler was manually computed and then I had to program a constexpr algorithm for it.
modm::platformnamespace.Rcc::LsiandRcc::Hsiclock.modm.digsepfilter for nicely formatting long integers.SystemClock::Iwdgto every board.assertBaudrateInToleranceimplementation.seconds_t,milliseconds_t,microseconds_tfor template parameters.percent_tcan be float since C++20.from_rangetofrom_linear.from_iteratortofrom_rangeand implement it to be compatible withstd::ranges::forward_range.