These functions make it easier to call the semihosting functions without
having to declare a local struct in each caller.
Since there is only one call with 4 indirect parameters, I did not add
a sys_semihost4 wrapper.

Optimize strcmp by calling orc.b (OR-Combine) instruction when Zbb extension
is supported instead of 4 instructions to detect if zero byte is exist in word

Signed-off-by: Mohamed Moawad <moawad@synopsys.com>

Enable zbb extension in CI for RISC-V to test optimizations using bit manipulation instructions like `orc.b`. 
This ensures that the strcmp optimization and other features using zbb are tested and validated

Signed-off-by: Mohamed Moawad <moawad@synopsys.com>

This is currently only implemented for Arm32 v7a where the CPU starts
off with MMU disabled. This requires 8KB of page table memory, so there
is an option to disable it if necessary.

Non-M processors have a wealth of stack pointers, one for each
interrupt entry point. As the only thing we're using the other entry
points for in the sample code is to catch errors, we'll just smash
the same stack for every one.

This allows us to stop setting the SP in the vectors themselves which
trashed one of the saved register values.

This is mixed in with code which uses ARM mode exception vectors on
'A' profile targets; attempting to set the TE bit on those doesn't
seem to work on QEMU.

Signed-off-by: Keith Packard <keithp@keithp.com>

We need to distinguish between v7, v7-a and v7-r targets as the v7-r
don't support the 32-bit ARM ISA and hence can't use ARM mode for
exception handling, while the v7-a only seem to support ARM mode for
exception handling.

Signed-off-by: Keith Packard <keithp@keithp.com>

…naligned

This seems like it got scrambled when merging newlib updates from arm.

Signed-off-by: Keith Packard <keithp@keithp.com>

copy-paste programming error left them all using REASON_UNDEF. Also
switch instruction to 'movs', which makes clang happy.

Signed-off-by: Keith Packard <keithp@keithp.com>

Clang doesn't let us do any regular C code from a naked function, so move
all of the Thumb-1 stack pointer setup into a single function and call that
from _cstart instead of _start.

Signed-off-by: Keith Packard <keithp@keithp.com>

clang sticks 'v5T' into the CPU_arch tag instead of 'v5TE', so check
for that value.

Signed-off-by: Keith Packard <keithp@keithp.com>

Allow these architectures to be tested

Signed-off-by: Keith Packard <keithp@keithp.com>

These fail on current qemu which now checks memory access alignment
when the MMU is disabled.

Signed-off-by: Keith Packard <keithp@keithp.com>

Make sure these two architectures work when using clang.

Signed-off-by: Keith Packard <keithp@keithp.com>

In this mode, longs are only 32 bits, so we need to use the 32-bit
form of the fcvtzs instruction instead of the 64-bit form.

Signed-off-by: Keith Packard <keithp@keithp.com>

The asm version was missing a check for LP64 causing it to also define
the function.

Signed-off-by: Keith Packard <keithp@keithp.com>

We need to enable the MMU so that unaligned access to normal memory
doesn't trap. On aarch64, that's not too bad as we can create level-1
block maps that cover 1GB apiece and then set up an 8GB virtual
space. That means we only need 8 PTEs.

Signed-off-by: Keith Packard <keithp@keithp.com>

This provides for override-able interrupt handling at the cost of
wasting all of the room provided for these functions in the interrupt
table.

Signed-off-by: Keith Packard <keithp@keithp.com>

Handle exceptions in semihost mode so that we'll get useful data.

Signed-off-by: Keith Packard <keithp@keithp.com>

Now that exceptions are trapped, make sure they work.

Signed-off-by: Keith Packard <keithp@keithp.com>

This uses the existing aarch64-zephyr cross compile file.

Signed-off-by: Keith Packard <keithp@keithp.com>

Now that the MMU is enabled, testing can be turned back on.

Signed-off-by: Keith Packard <keithp@keithp.com>

Update the Zephyr SDK and add aarch64 bits

Signed-off-by: Keith Packard <keithp@keithp.com>

…trtod()`

Quoting from C standard part 7.19.6.2

> `a,e,f,g` Matches an optionally signed floating-point number, infinity, or NaN, whose
format is the same as expected for the subject sequence of the `strtod`
function. The corresponding argument shall be a pointer to floating.

Regarding `strtod`, quoting from C standard part 7.20.1.3

> A character sequence `NAN` or `NAN(n-char-sequence)`, is interpreted as a quiet
NaN, if supported in the return type, else like a subject sequence part that does not have
the expected form; the meaning of the `n-char sequences` is implementation-defined.

Signed-off-by: Abdallah Abdelhafeez (Abdallahs70)

Make the new code match the old format

Signed-off-by: Keith Packard <keithp@keithp.com>

To avoid depending on a .specs file, we can't let meson insert the
libraries relative to the link_args by itself as the link_args end up
after the libs, which breaks the printf/scanf symbol definitions.

Signed-off-by: Keith Packard <keithp@keithp.com>

The generated specs files reference installed versions of the library
files.  If those are already present in the system, we may end up
using them for things like #include_next. The existing compile and
link args already provide all of the values which the .specs files
want to add.

Signed-off-by: Keith Packard <keithp@keithp.com>

I couldn't figure out why it got confused about the bounds checking
for pmatches.

Signed-off-by: Keith Packard <keithp@keithp.com>

Keep stdio.h from making other standard symbols visible

Signed-off-by: Keith Packard <keithp@keithp.com>

Follow POSIX requirements for visible symbols by placing them all in
signal.h and then cleaning that up to stop making other symbols
visible.

Signed-off-by: Keith Packard <keithp@keithp.com>

sys/cdefs already includes this

Signed-off-by: Keith Packard <keithp@keithp.com>

move 'error_t' to argz.h and envz.h where it belongs.
Get rid of __LINUX_ERRNO_EXTENSIONS__ -- C allows the library
to define as many errno values as it likes.

Signed-off-by: Keith Packard <keithp@keithp.com>

Move sys/cdefs include to the top of the file.

Signed-off-by: Keith Packard <keithp@keithp.com>

sys/config.h is included by sys/cdefs.h, which is included by
every public header.

Signed-off-by: Keith Packard <keithp@keithp.com>

This will be included by any public header, so the internal
source files don't need to also include it.

Signed-off-by: Keith Packard <keithp@keithp.com>

Not even newlib still supports any of this mechanism. Remove the
collate.c file and eliminate all internal usage of __collate_load_error.

Signed-off-by: Keith Packard <keithp@keithp.com>

PATH_MAX is a posix value, not a C value.

Signed-off-by: Keith Packard <keithp@keithp.com>

Make sure limits.h doesn't expose any non-C limits unless the
application requests them.

Signed-off-by: Keith Packard <keithp@keithp.com>

This is a machine-specific file, not an OS specific file.

Signed-off-by: Keith Packard <keithp@keithp.com>

This is declared in sys/_timespec.h already, no need to duplicate it
here.

Signed-off-by: Keith Packard <keithp@keithp.com>

These are required by C.

Signed-off-by: Keith Packard <keithp@keithp.com>

_POSIX2_RE_DUP_MAX is the minimum acceptable value for any POSIX
system while RE_DUP_MAX is the value supported by the library.

Signed-off-by: Keith Packard <keithp@keithp.com>

Adding an implementation of the bounds-checking C functions (as specified in Annex K of the C11 standard) to the PicoLibc.
These functions lower the risk of introducing security vulnerabilities such as buffer overflows and format string vulnerabilities into your code by providing clear and easy-to-use interfaces.
For each C function a secure alternate function ending in a "_s" postfix is provided (e.g., strcpy_s).
Use of these functions is recommended by security experts and secure coding standards.

also, Implemented unit tests for the Annex-K functions to ensure their corrctness.
Covered various scenarios including normal operation, boundary conditions, and error handling.

Signed-off-by: Mostafa Salman <mostafas@synopsys.com>