there were two problems:
* omitted underflow on subnormal results: exp2l(-16383.5) was calculated
as sqrt(2)*2^-16384, the last bits of sqrt(2) are zero so the down scaling
does not underflow eventhough the result is in subnormal range
* spurious underflow for subnormal inputs: exp2l(0x1p-16400) was evaluated
as f2xm1(x)+1 and f2xm1 raised underflow (because inexact subnormal result)

the first issue is fixed by raising underflow manually if x is in
(-32768,-16382] and not integer (x-0x1p63+0x1p63 != x)

the second issue is fixed by treating x in (-0x1p64,0x1p64) specially

for these fixes the special case handling was completely rewritten

only fma used these macros and the explicit union is clearer

* use new ldshape union consistently
* add ld128 support to frexpl
* simplify sqrtl comment (ld64 is not just arm)

remove STRICT_ASSIGN (c99 semantics is assumed) and use the conventional
union to prepare the scaling factor (so libm.h is no longer needed)

in lgammal don't handle 1 and 2 specially, in fma use the new ldshape
union instead of ld80 one.

* use float_t and double_t
* cleanup subnormal handling
* bithacks according to the new convention (ldshape for long double
and explicit unions for float and double)

* don't care about inexact flag
* use double_t and float_t (faster, smaller, more precise on x86)
* exp: underflow when result is zero or subnormal and not -inf
* exp2: underflow when result is zero or subnormal and not exact
* expm1: underflow when result is zero or subnormal
* expl: don't underflow on -inf
* exp2: fix incorrect comment
* expm1: simplify special case handling and overflow properly
* expm1: cleanup final scaling and fix negative left shift ub (twopk)

ld128 support was added to internal kernel functions (__cosl, __sinl,
__tanl, __rem_pio2l) from freebsd (not tested, but should be a good
start for when ld128 arch arrives)

__rem_pio2l had some code cleanup, the freebsd ld128 code seems to
gather the results of a large reduction with precision loss (fixed
the bug but a todo comment was added for later investigation)

the old copyright was removed from the non-kernel wrapper functions
(cosl, sinl, sincosl, tanl) since these are trivial and the interesting
parts and comments had been already rewritten.

* added ld128 support from freebsd fdlibm (untested)
* using new ldshape union instead of IEEEl2bits
* inexact status flag is not supported

method: if there is a large difference between the scale of x and y
then the larger magnitude dominates, otherwise reduce x,y so the
argument of sqrt (x*x+y*y) does not overflow or underflow and calculate
the argument precisely using exact multiplication. If the argument
has less error than 1/sqrt(2) ~ 0.7 ulp, then the result has less error
than 1 ulp in nearest rounding mode.

the original fdlibm method was the same, except it used bit hacks
instead of dekker-veltkamp algorithm, which is problematic for long
double where different representations are supported. (the new hypot
and hypotl code should be smaller and faster on 32bit cpu archs with
fast fpu), the new code behaves differently in non-nearest rounding,
but the error should be still less than 2ulps.

ld80 and ld128 are supported

* results are exact
* modfl follows truncl (raises inexact flag spuriously now)
* modf and modff only had cosmetic cleanup
* remainder is just a wrapper around remquo now
* using iterative shift+subtract for remquo and fmod
* ld80 and ld128 are supported as well

* faster, smaller, cleaner implementation than the bit hacks of fdlibm
* use arithmetics like y=(double)(x+0x1p52)-0x1p52, which is an integer
neighbor of x in all rounding modes (0<=x<0x1p52) and only use bithacks
when that's faster and smaller (for float it usually is)
* the code assumes standard excess precision handling for casts
* long double code supports both ld80 and ld128
* nearbyint is not changed (it is a wrapper around rint)

use -1/(x*x) instead of -1/(x+0) to return -inf, -0+0 is -0 in
downward rounding mode

* consistent code style
* explicit union instead of typedef for double and float bit access
* turn FENV_ACCESS ON to make 0/0.0f raise invalid flag
* (untested) ld128 version of ilogbl (used by logbl which has ld128 support)

new ldshape union, ld128 support is kept, code that used the old
ldshape union was rewritten (IEEEl2bits union of freebsd libm is
not touched yet)

ld80 __fpclassifyl no longer tries to handle invalid representation

switch to the new __block_all_sigs/__restore_sigs internal API to
clean up the code too.

this protects against deadlock from spurious signals (e.g. sent by
another process) arriving after the controlling thread releases the
other threads from the sync operation.

the head pointer was not being reset between calls to synccall, so any
use of this interface more than once would build the linked list
incorrectly, keeping the (now invalid) list nodes from the previous
call.

clone will pass the return value of the start function to SYS_exit
anyway; there's no need to call the syscall directly.

the wide variant was missed in the previous commit.

invalid format strings invoke undefined behavior, so this is not a
conformance issue, but it's nicer for scanf to report the error safely
instead of calling free on a potentially-uninitialized pointer or a
pointer to memory belonging to the caller.

rather than allocating a PATH_MAX-sized buffer when the caller does
not provide an output buffer, work first with a PATH_MAX-sized temp
buffer with automatic storage, and either copy it to the caller's
buffer or strdup it on success. this not only avoids massive memory
waste, but also avoids pulling in free (and thus the full malloc
implementation) unnecessarily in static programs.

this avoids failure if the file is not readable and avoids odd
behavior for device nodes, etc. on old kernels that lack O_PATH, the
old behavior (O_RDONLY) will naturally happen as the fallback.

DESTDIR was wrongly included in the symlink contents.

commit 07827d1a seems to have
introduced this issue. sigqueue is called from the synccall core, at
which time, even implementation-internal signals are blocked. however,
pthread_sigmask removes the implementation-internal signals from the
old mask before returning, so that a process which began life with
them blocked will not be able to save a signal mask that has them
blocked, possibly causing them to become re-blocked later. however,
this was causing sigqueue to unblock the implementation-internal
signals during synccall, leading to deadlock.

the BSD and GNU versions of this structure differ, so exposing it in
the default _BSD_SOURCE profile is possibly problematic. both versions
could be simultaneously supported with anonymous unions if needed in
the future, but for now, just omitting it except under _GNU_SOURCE
should be safe.

I originally added this warning option based on a misunderstanding of
how it works. it does not warn whenever the destination of the cast
has stricter alignment; it only warns in cases where misaligned
dereference could lead to a fault. thus, it's essentially a no-op for
i386, which had me wrongly believing the code was clean for this
warning level. on other archs, numerous diagnostic messages are
produced, and all of them are false-positives, so it's better just not
to use it.

unlike the old C memcpy, this version handles word-at-a-time reads and
writes even for misaligned copies. it does not require that the cpu
support misaligned accesses; instead, it performs bit shifts to
realign the bytes for the destination.

essentially, this is the C version of the ARM assembly language
memcpy. the ideas are all the same, and it should perform well on any
arch with a decent number of general-purpose registers that has a
barrel shift operation. since the barrel shifter is an optional cpu
feature on microblaze, it may be desirable to provide an alternate asm
implementation on microblaze, but otherwise the C code provides a
competitive implementation for "generic risc-y" cpu archs that should
alleviate the urgent need for arch-specific memcpy asm.

while the incorporation of this requirement from C99 into C++11 was
likely an accident, some software expects it to be defined, and it
doesn't hurt. if the requirement is removed, then presumably
__bool_true_false_are_defined would just be in the implementation
namespace and thus defining it would still be legal.

there is no non-dot version of the andis instruction, but there's no
harm in updating the flags anyway, so just use the dot version.

this version of memset is optimized both for small and large values of
n, and makes no misaligned writes, so it is usable (and near-optimal)
on all archs. it is capable of filling up to 52 or 56 bytes without
entering a loop and with at most 7 branches, all of which can be fully
predicted if memset is called multiple times with the same size.

it also uses the attribute extension to inform the compiler that it is
violating the aliasing rules, unlike the previous code which simply
assumed it was safe to violate the aliasing rules since translation
unit boundaries hide the violations from the compiler. for non-GNUC
compilers, 100% portable fallback code in the form of a naive loop is
provided. I intend to eventually apply this approach to all of the
string/memory functions which are doing word-at-a-time accesses.