The old code used to do OSMO_ASSERT(), i.e. checking if the
endpoint number was != 0. However, now we're checking if it's
!= 0xFF, as the default initialization is 0xFF, and an unset
endpoint can be detected this way.
Change-Id: I7d334724a4c6b89ef8faa49bed49ed57ea216ff9
undef syms from the left, def syms from the right, as it should be.
It's either this, or -Wl,--start-group/-Wl,--end-group around
everything.
Change-Id: I8083c5dc990acd6b60b053883fdd1b42c3bc0d4f
AFAIK these are not used yet, but have been added to the upstream
ASFv4 library.
I added them to avoid future false negative when diffing files.
Change-Id: Icf6fb641050a1e7132ff2b14c1ce537ec9403b9f
upstream ASFv4 updated the copyright year.
somehow they don't know you should use the range instead of the
last year :/
this change is just so I can't continue diffing ASF files without
false positive.
Change-Id: Ic90e692815b30921bc8750c240ba0a5d3f4ccc70
this add the documentation and (yet unused) RTC definitions so to
by in sync with the ASFv4 upstream library
Change-Id: If3e24ede24b4e8cc2724a796756a5a2aa81119de
With all the tons of ASF4 library code, ARM/Microchip couldn't even
bothered to implement the most basic CMSIS bits such as the
SystemCoreClockUpdate() function that would dynamically compute
the current SysTick rate at runtime.
To make things worse, they're also not using the #define they have
for the initial core clock rate that's set up at boot time but rather
have a hardcoded 48 MHz value as SystemCoreClock.
Without a correct SystemCoreClock, the jiffies rate is wrong, and
all related timing off.
Change-Id: I9966de4d9ad4e623893044fc341b9852659e8f39
We simply use the SysTick timer to count at 1KHz and fill a 64bit jiffies
conunter. This counter is then used to fill tv_sec and tv_usec in the
gettimeofday() implementation.
NOTE: tv_sec will not indicate the seconds since the epoch (Jan 01,
1970), but rather since system startup. For the existign users,
particularly osmo_timer, this doesn't matter.
Change-Id: I9dbbb730996bde1e7039f790d76d7243739a8419
This will display the time; However, as the host is not telling us
anything about the current date/time [yet], we will always be in
"N seconds since 1970".
Change-Id: Ic23152db65a01eefd28d67e0132a82352875e467
I really hate the Atmel copy+paste style makefile where every include path
is specified three times, rather than using variables.
Change-Id: I39dd880ced341aaea54e30e8b670dbaea445aa8b
We do so using our own 'raw' log target to avoid the 4k-on-stack
buffer of libosmocore _output() function.
Change-Id: I7a10b5b2b50bcee0154a1fa3fc43756aec836226
The IRQ output of the NCN8025 doesn't really have an "active" state
but it's slightly more intricate. To avoid confusion, let's just
provide a function to obtain the raw value (0=low / 1=high) and
let the user decide what to do with that
Change-Id: I752a84c67046b1d5f832cd1ef2070ec9b2489231
Harald Welte
Eric Wild
Kevin Redon