| Commit message (Collapse) | Author | Age |
|
|
|
| |
This is much less work and actually works better.
|
|
|
|
| |
com.foo.bar instead of com/foo/bar, only convert to latter form when needed.
|
|
|
|
|
| |
Boolean array region was missed by mistake.
Too lazy to split this commit so yeah, included here..
|
|
|
|
|
| |
Makes sure stuff appends even if LDFLAGS etc are provided on command
line.
|
|
|
|
|
| |
bionic uses different sized types, also bugs:
https://android.googlesource.com/platform/bionic/+/master/docs/32-bit-abi.md
|
| |
|
|
|
|
| |
Contains some unity specific java functions
|
|
|
|
| |
Mostly the -DWOLF define to switch between apks more easily
|
| |
|
|
|
|
|
|
|
| |
Real Android has lower limit, (IIRC the 256 we had), but since
we don't actually do any ref counting yet, nothing is ever unloaded.
Thus lets just raise this limit for now.
|
| |
|
|
|
|
| |
Wolf simulator (lol) starts working
|
|
|
|
|
| |
Tries call method from generic java object or class if implementation
is not found for the topmost level.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
|
|
| |
Instead of creating class, return the method instead.
Check for type mismatch in functions that use these objects in our jvm
implementation, but on mismatch print the mismatch, but return a empty
object (all-0) and stuff could still work. Essentially effect is having
container/object with no data.
|
|
|
|
|
| |
For some reason doesn't work with NULL act, need to investigate more
later.
|
| |
|
|
|
|
|
| |
Currently doesn't do anything different from jvm_get_object, but can be
used to log cases where unexpected object was returned.
|
|
|
|
|
| |
This is better than returning NULL, as we can track better where things
go wrong and what needs to be implemented.
|
| |
|
| |
|
| |
|
|
|
|
|
| |
No need to give runtime dependency for everything anymore, just to have
a directory.
|
| |
|
| |
|
|
|
|
| |
Used in android for stack unwinding for c++
|
| |
|
| |
|
|
|
|
|
| |
Supporting host libstdc++ is going to be complicated.
Because c++'s ABI is unstable and mess.
|
| |
|
| |
|
|
|
|
|
| |
We want to use android's libstdc++, so we need to pass __cxa_demangle
from linker.
|
|
|
|
| |
Casting symbol is wrong.
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
| |
This is why the bionic_file_to_glibc_file did not work with just
pointers. The input is actually pointer to our wrapper, doh.
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Bionic libc does stupid stuff and defines standard streams as macros to
reference of an array element. Thus we can't just point standard streams
to their respective correct addresses. (Why does both glibc and bionic
even expose the internal FILE structs??)
So we have to wrap every function that takes FILE* as argument and
mangle it with bionic_file_to_glibc_file. We can't just check addresses
because gnu's libstdc++ copies the FILE*.
I assume same thing happens with the c++ locale support, and we need to
do more complex ctype wrapping.
|
|
|
|
|
|
|
|
| |
Remove eglGetDisplay as it's called with NULL anyways
(Reimplement if otherwise in some app)
In eglCreateWindowSurface, check that our pointer really is
ANativeWindow, otherwise try use it as is.
|
|
|
|
| |
Thanks gcc's printf attribute
|
|
|
|
|
| |
Static header implementation is not thread safe as mutex is of course
duplicated to each version...
|
| |
|
|
|
|
| |
Allows us to return "valid" pointer when bionic program calls dlopen.
|
| |
|
| |
|