Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

get_node already adds a ref.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Working around a Rosetta bug on Apple Silicon - faults for certain
instructions (e.g. lock cmpxchg8b) are misreported, which can break
VPROT_WRITEWATCH handling.

The virtual_monitor used for non-interactive service executions do not have
an adapter set. Avoid segfaulting and poisoning the mutex in such cases.

Saving a huge number of .refptr indirections for all extern symbols.

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

This is needed for duplicated argument names, as the last arg will shadow
all the prior ones when it comes to name lookup.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Rather than using the variable obj for it. Since it's an implementation
detail, the scope's dispex object accesses them using index props (using
same indices as the buffer's).
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

It was confusing and aliased to obj when it was a jsdisp (and shared ref), but
we can obtain that already with helpers as needed (as_jsdisp and to_jsdisp),
no reason to keep it so confusing and a separate field.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

`ref` can be negative in case it refers to an argument. Even though scope !=
frame->base_scope would rule this out (because only base scopes have args),
it was checked *after* the memory access, which would read out of bounds
memory first. This didn't appear as an issue in practice since it's using the
heap pool, so there's probably valid memory before it, but it's still wrong.
Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Signed-off-by: Gabriel Ivăncescu <gabrielopcode@gmail.com>

Previously if an empty string was encountered while parsing out PATH,
cmd.exe would ignore the rest of the PATH entirely.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55097Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55097Signed-off-by: Liam Middlebrook <lmiddlebrook@nvidia.com>

This is probably a leftover of when Vulkan swapchain images were
used directly.

The original ntdll may have been hooked.

realloc may free lead_static.

Based on a patch from Byeong-Sik Jeon <bsjeon@hanmail.net>.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55027

The composition or result strings may be present, but with zero length.
In which case we still want to send the messages, to indicate any change
for instance whenever a character is deleted and strings become empty.

Wine-Bug: https://bugs.winehq.org/show_bug.cgi?id=55055