some notes:
- library search path is hard coded
- x86_64 code is untested and may not work
- dlopen/dlsym is not yet implemented
- relocations in read-only memory won't work

this seems to be necessary to make the linker accept the functions in
a shared library (perhaps to generate PLT entries?)

strictly speaking libc-internal asm should not need it. i might clean
that up later.

if thread id was reused by the kernel between the time pthread_kill
read it from the userspace pthread_t object and the time of the tgkill
syscall, a signal could be sent to the wrong thread. the tgkill
syscall was supposed to prevent this race (versus the old tkill
syscall) but it can't; it can only help in the case where the tid is
reused in a different process, but not when the tid is reused in the
same process.

the only solution i can see is an extra lock to prevent threads from
exiting while another thread is trying to pthread_kill them. it should
be very very cheap in the non-contended case.

previously a long-running dtor could cause pthread_detach to block.

these are useless and have caused problems for users trying to build
with non-gnu tools like tcc's assembler.

at present the i386 code does not support sse floating point, which is
not part of the standard i386 abi. while it may be desirable to
support it later, doing so will reduce performance and require some
tricks to probe if sse support is present.

this first commit is i386-only, but it should be trivial to port the
asm to x86_64.

even if size_t was 32-bit already, the fact that the value was
unsigned and that gcc is too stupid to figure out it would be positive
as a signed quantity (due to the immediately-prior arithmetic and
conditionals) results in gcc compiling the integer-to-float conversion
as zero extension to 64 bits followed by an "fildll" (64 bit)
instruction rather than a simple "fildl" (32 bit) instruction on x86.
reportedly fildll is very slow on certain p4-class machines; even if
not, the new code is slightly smaller.

looks like busybox is going to want it, and apparently some other
low-level network software does too...

unfortunately traditional i386 practice was to use "long" rather than
"int" for wchar_t, despite the latter being much more natural and
logical. we followed this practice, but it seems some compilers (clang
and maybe certain gcc builds or others too..?) have switched to using
int, resulting in spurious pointer type mismatches when L"..." wide
strings are used. the best solution I could find is to use the
compiler's definition of wchar_t if it exists, and otherwise fallback
to the traditional definition.

there's no point in duplicating this approach on 64-bit archs, as
their only 32-bit type is int.

this is not too ugly and should result in significant code size and
performance improvements for many programs.

basically there are 3 choices for how to implement this variable-size
string member:
1. C99 flexible array member: breaks using dirent.h with pre-C99 compiler.
2. old way: length-1 string: generates array bounds warnings in caller.
3. new way: length-NAME_MAX string. no problems, simplifies all code.

of course the usable part in the pointer returned by readdir might be
shorter than NAME_MAX+1 bytes, but that is allowed by the standard and
doesn't hurt anything.

this actually inadvertently disallows some valid patterns with
redundant / or * characters, but it's better than allowing unbounded
vla allocation.

eventually i'll write code to move the pattern to the stack and
eliminate redundancy to ensure that it fits in PATH_MAX at the
beginning of glob. this would also allow it to be modified in place
for passing to fnmatch rather than copied at each level of recursion.

there is a resource limit of 0 bits to store the concurrency level
requested. thus any positive level exceeds a resource limit, resulting
in EAGAIN. :-)