Apply mandatory 2-space indentation for YAML arrays in
spec/build/testsuites/sptests/grp.yml to satisfy the RTEMS CI
formatting checks.
No functional change.
Signed-off-by: Karthikey Kadati <karthikey3608@gmail.com>
Cast the literal 1 to rtems_error_code_t before shifting to avoid
undefined behavior on systems with a 16-bit int.
Include <limits.h> for CHAR_BIT.
Signed-off-by: Shivam Deolankar <shivam.deolankar@gmail.com>
These changes eliminate the year 2100 as a limit for TOD. The new
limit is the year 4096. This limit is not due to any RTEMS limitation
but just before the year 4183 when the rounding error in the length
of the year adds up to an entire day. See
https://de.wikipedia.org/wiki/Schaltjahr for details.
There are still limits on Classic APIs which use the 32-bit rtems_interval
type. This corresponds to the Score Watchdog_Interval type. This impacts
delays, timeouts, and Classic API "after" timers.
Closes#5143
Co-authored-by: Joel Sherrill <joel@rtems.org>
Co-authored-by: Gedare Bloom <gedare@rtems.org>
Newlib hash: a7c614986ab2e21a342afe3105fb88985a653996
These were added to allow cpukit/score code to use mktime() and
gmtime_r() while avoiding dependencies on external libraries.
The code seems to be broken for years.
Tested to work on real hardware (AMD Athlon) booted
over pxelinux in preparation for CAN stack test
with original Philips/NXP SJA1000 chips.
RTEMS configured with with
RTEMS_POSIX_API = True
BSP_ENABLE_COM1_COM4 = True
USE_COM1_AS_CONSOLE = False
USE_VBE_RM = False
USE_CIRRUS_GD5446 = False
Signed-off-by: Pavel Pisa <ppisa@pikron.com>
CID 1592923: Uninitialized scalar variable.
The 'spolicy' variable in thread::start_thread was declared without
initialization. Although the subsequent switch statement assigns
a value, Coverity flagged this as a risk.
This patch initializes 'spolicy' to SCHED_FIFO at declaration.
SCHED_FIFO was chosen because it matches the existing 'default'
case in the switch statement, ensuring consistency.
cpukit/librtemscxx: Initialize spolicy in start_thread (CID 1592923)
Restore default block for switch statement consistency
This adds a BSP-level function that returns the calculated PSCI
target_cpu parameter for a given cpu_index. BSPs can override this
function to use the correct affinity level for the calculation.
Also updates _CPU_SMP_Start_processor() to make use of this
function and remove code duplication.
963ba98047 introduced a type cast
conversion that causes a signed comparison instead of unsigned
that is always true. The checks for values in excess of
SCHEDULER_EDF_PRIO_MSB are not needed since this will be LONG_MIN.
We only need to make sure the budget and deadline are positive
non-zero values.
This is initial mvp implementation for the Pi5 BSP. The
initial support has the following features:
- PL011 debug uart (full termios console driver)
- Arm Generic Timer (virtual interface)
- GIC-V2 interrupts
This will be extended in follow-up work.
Should Close rtems/programs/gsoc#85
Resolving warnings on 3rd-party code can be invasive and will make it
harder to bring in updates from upstream. For issues outside -Wall,
disable the warning for this piece of code instead.
On 32bit systems, int and long are the same size. This is not the case
on 64bit systems. This corrects the maximum value of the stringto type
on those systems.
The function cast from more parameters to fewer parameters is generally
unsafe as the additional parameters will be undefined values that happen
to be in the right registers if it is ever called. Instead of casting
the function pointer, provide a dummy function pointer when the original
function pointer is non-NULL since it is only ever used for a
NULL-check.
libc's strto* functions' behaviour isn't defined when the passed string
is NULL. In glibc it causes a seg fault when I tested it. This could
cause inconsistency.
rtem's stringto* functions didnt previously checked for null before
passing to underlying strto* functions.
Added the check which returns `RTEMS_INVALID_ADDRESS` on failure.
Also added a test case for the same
Signed-off-by: Prashant Rahul <prashantrahul141@protonmail.com>
This warning occurs when comparing a signed variable to an unsigned one.
This addresses warnings that only occurred on 64-bit targets. For the
ones which only appeared on 64-bit targets, the cause was frequently
a mismatch when comparing a combination off_t, ssize_t, and int.