Kernel syscall API no longer zeros all unused argument registers and
libc now uses inlined syscall macro internally. This significantly
cleans up generated code for basic syscall wrapper functions.
We were linking with -nostdlib and manually linked against libgcc. This
does not link with crtbegin and crtend which provides __dso_handle
preventing use of some global C++ constructors inside libc.
Now we just don't link against libc fixing this issue
There were two problems with my previous implementation
- TLS was not allocated if nothing used it. There is a fallback
initialization in _init_libc, but this was not enough if one of the
init functions tried to access errno.
- __tls_get_addr was not resolved. If __tls_get_addr was called through
a plt entry, everything would just break :(
This implementation is on top of inodes instead of fds as linux does it.
If I start finding ports/software that relies on epoll allowing
duplicate inodes, I will do what linux does.
I'm probably missing multiple epoll_notify's which may cause hangs but
the system seems to work fine :dd:
I now set library SONAME so i don't have to build them in build root.
Creating font now makes sure the install directory exists. This allows
building using make files.
LibC now links agains ligcc
Base sysroot is not only installed when creating the syroot directory
for the first time. To get it installed again, you have to either delete
the old sysroot diretory or run clean target
This makes it so its actually possible to edit files specified in the
base sysroot without random hacks.
SSE is now unconditionally enabled any where and most of math.h is now
actually implemented. using __builtin_<func> lead to many hangs where
the builtin function would just call itself.
some ports want to link against these, but they are not ever called
based on simple testing
This patch adds stubs for
- openlog
- syslog
- getrusage
- mlock
- mprotect
- getpeername
- shutdown
- tzset
- mktime
