patch by nsz

dynamic-allocation of the structure is not valid; it can crash an
application if malloc fails. since localeconv is not specified to have
failure conditions, the object needs to have static storage duration.

need to review whether all the values are right or not still..

this was actually dangerously wrong, but presumably nobody uses this
broken function anymore anyway..

if we eventually have build options, it might be nice to make an
option to dummy this out again, in case anybody needs a system-wide
disable for disk/ssd-thrashing, etc. that some daemons do when
logging...

large precision values could cause out-of-bounds pointer arithmetic in
computing the precision cutoff (used to avoid expensive long-precision
arithmetic when the result will be discarded). per the C standard,
this is undefined behavior. one would expect that it works anyway, and
in fact it did in most real-world cases, but it was randomly
(depending on aslr) crashing in i386 binaries running on x86_64
kernels. this is because linux puts the userspace stack near 4GB
(instead of near 3GB) when the kernel is 64-bit, leading to the
out-of-bounds pointer arithmetic overflowing past the end of address
space and giving a very low pointer value, which then compared lower
than a pointer it should have been higher than.

the new code rearranges the arithmetic so that no overflow can occur.

while this bug could crash printf with memory corruption, it's
unlikely to have security impact in real-world applications since the
ability to provide an extremely large field precision value under
attacker-control is required to trigger the bug.

request/patch by william haddonthethird, slightly modifed to add
_GNU_SOURCE feature test macro so that the compiler can verify the
prototype matches.

this is mildly ugly, but less ugly than gnulib trying to poke at the
definition of the FILE structure...

for seekable files, posix imposed requirements on the offset of the
underlying open file description after a stream is closed. this was
correctly handled (as a side effect of the unconditional fflush call)
when streams were explicitly closed by fclose, but was not handled
correctly at program exit time, where fflush(0) was being used.

the weak symbol hackery is to pull in __stdio_exit if either of
__toread or __towrite is used, but avoid calling it twice so we don't
have to keep extra state. the new __stdio_exit is a streamlined fflush
variant that avoids performing any unnecessary operations and which
never unlocks the files or open file list, so we can be sure no other
threads write new data to a stream's buffer after it's already
flushed.

there is no need/use for a flush hook. the write function serves this
purpose already. i originally created the hook for implementing mem
streams based on a mistaken reading of posix, and later realized it
wasn't useful but never removed it until now.

apparently this was never tested before.

the old behavior was to only consider a stream to be "reading" or
"writing" if it had buffered, unread/unwritten data. this reportedly
differs from the traditional behavior of these functions, which is
essentially to return true as much as possible without creating the
possibility that both __freading and __fwriting could return true.

gnulib expects __fwriting to return true as soon as a file is opened
write-only, and possibly expects other cases that depend on the
traditional behavior. and since these functions exist mostly for
gnulib (does anything else use them??), they should match the expected
behavior to avoid even more ugly hacks and workarounds...

one file was reusing another file's macro name, and many had
inconsistent underscores and application of SYS prefix, etc.

patch by Szabolcs Nagy (nsz)

it probably does not matter for /dev/null, but this should be done
consistently anyway.

this is required in case dtors use stdio.

also remove the old comments; one was cruft from when the code used to
be using function pointers and conditional calls, and has little
motivation now that we're using weak symbols. the other was just
complaining about having to support dtors even though the cost was
made essentially zero in the non-use case by the way it's done here.

stime is not _XOPEN_SOURCE, and some functions were missing with
_BSD_SOURCE..

these are not exposed publicly in any header, but the few programs
that use them (modutils/kmod, etc.) are declaring the functions
themselves rather than making the syscalls directly, and it doesn't
really hurt to have them (same as the capset junk).

based on patch by Emil Renner Berthing, with minor changes to dirent.h
for LFS64 and organization of declarations

this code should work unmodified once a real strverscmp is added, but
I've been hesitant to add it because the GNU strverscmp behavior is
harmful in a lot of cases (for instance if you have numeric filenames
in hex). at some point I plan on trying to design a variant of the
algorithm that behaves better on a mix of filename styles.

these were left in glibc for binary compatibility after the public
part of the interface was removed, and libcap kept using them (with
its own copy of the header files) rather than just making the syscalls
directly. might as well add them since they're so small...

i originally omitted these (optional, per POSIX) interfaces because i
considered them backwards implementation details. however, someone
later brought to my attention a fairly legitimate use case: allocating
thread stacks in memory that's setup for sharing and/or fast transfer
between CPU and GPU so that the thread can move data to a GPU directly
from automatic-storage buffers without having to go through additional
buffer copies.

perhaps there are other situations in which these interfaces are
useful too.

the types of these expressions must match the integer promotions.
unsigned 8- and 16-bit values promote to signed int, not unsigned int.

signedness issue kept %ls with no precision from working at all

printf was not printing too many characters, but it was reading one
too many wchar_t elements from the input. this could lead to crashes
if running off the page, or spurious failure if the conversion of the
extra wchar_t resulted in EILSEQ.

this broke the busybox "free" utility (memory reporting) and possibly
other things like uptime.

the field width limit was not being cleared before reading the
literal, causing spurious failures in scanf in cases like "%2d:"
scanning "00:".

this issue affects the last gpl2 version of binutils, which some
people are still using out of aversion to gpl3. musl requires
-Bsymbolic-functions because it's the only way to make a libc.so
that's able to operate prior to dynamic linking but that still behaves
correctly with respect to global vars that may be moved to the main
program via copy relocations.

it's possible that the user has provided a compiler that does not have
any libc to link to, so linking a main program is a bad idea. instead,
generate an empty shared library with no dependencies.

in theory we could support stack protector in the libc itself, and
users wanting to experiment with such usage could add
-fstack-protector to CFLAGS intentionally. but to avoid breakage in
the default case, override broken distro-patched gcc that forces stack
protector on.

some broken distro-provided toolchains have modified gcc to produce
only "gnu hash" dynamic hash table by default. as this is unsupported
by musl, that results in a non-working libc.so. we detect and switch
this on in configure rather than hard-coding it in the Makefile
because it's not supported by old binutils versions, but that might
not even be relevant since old binutils versions already fail from
-Bsymbolic-functions being missing. at some point I may review whether
this should just go in the Makefile...