The code accessing the floating point control/status register, namely
#define __cfc1(__fcsr) __asm __volatile("cfc1 %0, $31" : "=r" (__fcsr)
does not compile with mips16. This changed the makefile to pass -mno-mips16 to avoid the following
compiler error:
mips-mti-elf fails with "Error: unrecognized opcode `cfc1 $3,$31'"
- Currently, printf("%La\n", 1e1000L) crashes with segv due to lack
of frexpl() function. With this patch, frexpl() function has been
implemented in libm to solve this issue.
Addresses: https://sourceware.org/pipermail/newlib/2021/018718.html
- If the number has large integer part and small fraction part is
specified in output format, e.g. printf("%.3f", sqrt(2)*1e60);,
valid output digits were insufficient. This patch fixes the issue.
https://cygwin.com/pipermail/cygwin/2021-November/249930.html
reported a regression introduce by using a dynamically sized local
char array in favor of a statically sized array.
Fix this by reverting to a statically sized array, using a small
buffer on the stack for a reasonable number of requested digits, a
big mallocated buffer otherwise. This should work for small targets
as well, given that malloc is used in printf anyway right now.
This is *still* hopefully just a temporary measure, unless somebody
actually provides a new ldtoa.
Fixes: 4d90e53359 ("ldtoa: fix dropping too many digits from output")
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Use the same name as glibc & gnulib to indicate "newlib itself is
being compiled". This also harmonizes the codebase a bit in that
_LIBC was already used in places instead of _COMPILING_NEWLIB.
Building for bfin-elf, mips-elf, and x86_64-pc-cygwin produces
the same object code.
Some distros enable _FORTIFY_SOURCE by default which upsets building
newlib which itself implements the logic for this define. For example,
building gets.c fails because the includes set up a gets() macro which
expands in the definition.
Since newlib isn't prepared to build itself with _FORTIFY_SOURCE, and
it's not clear if it's even useful, ignore it when building the code.
This also matches what glibc is doing.
The _COMPILING_NEWLIB symbol is for declaring "the code is being
compiled for newlib itself" so headers can change behavior vs the
header being used by users (who should get the normal clean API).
Unfortunately, this symbol is defined inconsistently leading to it
only being useful for a few subsections of the tree.
Pull it out so that it's defined all the time for all targets.
We're seeing a build failure in GNU sim code which is using _P locally
but the ctype.h define clashes with it. Rename these to use the same
symbols that glibc does. They're a bit more verbose, but seems likely
that we'll have fewer conflicts if glibc isn't seeing them.
However, these shortnames are still used internally by ctype modules
to produce pretty concise source code, so move the short names to the
internal ctype_.h where short name conflicts shouldn't show up.
This code looks like it's written to be copied & pasted between diff
C libraries and relies on _LIBC only being used with glibc. This will
break when newlib changes from _COMPILING_NEWLIB to _LIBC, so delete
the glibc-specific logic ahead of time.
ldtoa cuts the number of digits it returns based on a computation of
number of supported bits (144) divide by log10(2). Not only is the
integer approximation of log10(2) ~= 8/27 missing a digit here, it
also fails to take really small double and long double values into
account.
Allow for the full potential precision of long double values. At the
same time, change the local string array allocation to request only as
much bytes as necessary to support the caller-requested number of
digits, to keep the stack size low on small targets.
In the long run a better fix would be to switch to gdtoa, as the BSD
variants, as well as Mingw64 do.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Currently, newlib does not declare strsignal if DEFS_H is defined,
ostensibly to work around a gdb bug. However, gdb itself compiles
even with this ifndef removed, and this makes sim (another part of
gdb) fail to compile.
Since it is not clear exactly what issue this was working around,
this patch just replaces that ifdef with the correct check,
i.e. __POSIX_VISIBLE >= 200809.
Reported by prodisDown:
In picolibc/newlib/libc/string/strrchr.c
if (i) { while ((s=strchr(s, i))) { last = s; s++; } } else { last = strchr(s, i); }
Value (for example 0xFFFFFF00) in if (i) can pass test and
then be typecasted to char inside strchr(). Then s++ and then
buffer overrun.
It can be fixed by preventive typecast i = (int) (char) i; or
typecasting inside expression if ((char) i).
Fixed by casting to char.
Signed-off-by: Keith Packard <keithp@keithp.com>
Add specialized rotations RB_RED_ROTATE_LEFT() and RB_RED_ROTATE_RIGHT() which
may be used if we rotate a red child which has a black sibling. Such a red
node must have at least two child nodes so that the following red-black tree
invariant is fulfilled:
Every path from a given node to any of its descendant NULL nodes goes through
the same number of black nodes.
PARENT
/ \
BLACK RED
/ \
BLACK BLACK
Add specialized rotations RB_PARENT_ROTATE_LEFT() and RB_PARENT_ROTATE_RIGHT()
which may be used if the parent node exists and the direction of the child is
known. The specialized rotations are derived from RB_ROTATE_LEFT() and
RB_ROTATE_RIGHT() where the RB_SWAP_CHILD() was replaced by a simple
assignment.
In RB_GENERATE_REMOVE_COLOR() simplify a chain of conditions of the following
pattern
if (x) {
...
} else if (!x) {
...
}
to
if (x) {
...
} else {
...
}
We have
#define RB_ISRED(elm, field) \
((elm) != NULL && RB_COLOR(elm, field) == RB_RED)
So, the RB_ISRED() contains an implicit check for NULL. In
RB_GENERATE_REMOVE_COLOR() the "elm" pointer cannot be NULL in the while
condition. Use RB_COLOR(elm) == RB_BLACK instead.
There are a few files that are tied to the GNU simulator, so add myself
to match the general sim project.
Also add myself to the general write-after-approval since I've been doing
that for a long time now anyways :).
When newlib is configured with --enable-newlib-reent-check-verify,
the assert macro is already defined in the nano-mallocr.c compile unit.
Contributed by STMicroelectronics
Signed-off-by: Torbjörn SVENSSON <torbjorn.svensson@st.com>
libm/machine/i386/f_ldexp.S:30: Warning: no instruction mnemonic suffix given and no register operands; using default for `fild'
libm/machine/i386/f_ldexpf.S:30: Warning: no instruction mnemonic suffix given and no register operands; using default for `fild'
fix this by adding the l mnemonic suffix
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
unused function warning for two_way_short_needle,
different char type warnings for standard string functions
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
_STDIO_WITH_THREAD_CANCELLATION_SUPPORT was never defined.
Include ../stdio/local.h to get the right definition per target.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
__STDINT_EXP() is provided by newlib but not by stdint-gcc.h. stdint-gcc.h
is used when the GCC argument -ffreestanding is used and this results in this
file not compiling.
This patch to the libc/machine/nvptx port of newlib implements an
approximation of "clock" and provides some additional stub routines.
These changes not only reduce the number of (link) failures in the GCC
testsuite when targeting nvptx-none, but also allow the NIST scimark4
benchmark to compile and run without modification.
newlib already contains support for backends to provide their own
clock implementations via -DCLOCK_PROVIDED. That functionality is
used here to return an approximate elapsed time based on the NVidia
GPU's clock64 cycle counter. Although not great, this is better than
the current behaviour of link error from the unresolved symbol
_times_r.
The other part of the patch is to add a small number of stub functions
to nvptx's misc.c. Adding isatty, for example, resolves linking
problems in libc from the dependency in __smakebuf_r, and the sync
stub, for example, fixes the failure with GCC's
testsuite/gfortran.dg/ISO_Fortran_binding_14.f90 [which simply tests
that gfortran can call a/any C function].
newlib/
configure.host: Add -DCLOCK_PROVIDED to newlib_cflags on nvptx*.
newlib/libc/machine/nvptx
Makefile.am: Add clock.c to lib_a_SOURCES.
clock.c: New source file to implement/approximate clock().
misc.c: Add stubs for fstat, isatty, open, sync and unlink.
_strtod_l as well as the gethex function both fetch the decimal point
from the current LC_NUMERIC locale info. This pulls in _C_numeric_locale
unconditionally even on targets not supporting locales at all.
Another problem is that strtod.c and gdtoa-gethex.c are ELIX 1, while
locale information in general isn't. This leads to potential build
breakage on bare metal targets.
Fix this by setting the decimal point to "." on all targets not
defining __HAVE_LOCALE_INFO__.
While at it, const'ify the entire local decimal point info in the
affected functions.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
svfwscanf replaces getwc and ungetwc_r. The comments in the code talk
about avoiding file operations, but they also need to bypass the
mbtowc calls as svfwscanf operates on wchar_t, not multibyte data,
which is a more important reason here; they would not work correctly
otherwise.
The ungetwc replacement has code which uses the 3 byte FILE _ubuf
field, but if wchar_t is 32-bits, this field is not large enough to
hold even one wchar_t value. Building in this mode generates warnings
about array overflow:
In file included from ../../newlib/libc/stdio/svfiwscanf.c:35:
../../newlib/libc/stdio/vfwscanf.c: In function '_sungetwc_r.isra':
../../newlib/libc/stdio/vfwscanf.c:316:12: warning: array subscript 4294967295 is above array bounds of 'unsigned char[3]' [-Warray-bounds]
316 | fp->_p = &fp->_ubuf[sizeof (fp->_ubuf) - sizeof (wchar_t)];
| ^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
In file included from ../../newlib/libc/stdio/stdio.h:46,
from ../../newlib/libc/stdio/vfwscanf.c:82,
from ../../newlib/libc/stdio/svfiwscanf.c:35:
../../newlib/libc/include/sys/reent.h:216:17: note: while referencing '_ubuf'
216 | unsigned char _ubuf[3]; /* guarantee an ungetc() buffer */
| ^~~~~
However, the vfwscanf code *never* ungets data before the start of the
scanning operation, and *always* ungets data which matches the input
at that point, so the code always hits the block which backs up over
the input data and never hits the block which uses the _ubuf field.
In addition, the svfwscanf code will always start with the unget
buffer empty, so the ungetwc replacement never needs to support an
unget buffer at all.
Simplify the code by removing support for everything other than
backing up over the input data, leaving the check to make sure it
doesn't get underflowed in case the vfscanf code has a bug in it.
Signed-off-by: Keith Packard <keithp@keithp.com>
Added function prototypes to newlib/libc/include/pthread.h
for the following Issue 8 Standard APIs:
pthread_cond_clockwait()
pthread_mutex_clocklock()
pthread_rwlock_clockrdlock()
pthread_rwlock_clockwrlock()
A recent patch introduced new code for sig2str/str2sig.
This code does not properly exclude code that requires
SIGRTMIN/SIGRTMAX to be defined and triggers the following
compile error:
newlib/libc/signal/sig2str.c:199:8: error: 'SIGRTMIN' undeclared
newlib/libc/signal/sig2str.c:200:29: error: 'SIGRTMAX' undeclared
Let's add the missing guards.
Fixes: 2b50ec0cd2 ("libc: Fix compilation for new sig2str/str2sig implementation")
Signed-off-by: Christoph Muellner <cmuellner@gcc.gnu.org>
Added implementations for sig2str() and str2sig() in libc/signal
in order to improve POSIX compliance. Added fucntion prototypes
in libc/include/sys/signal.h.
riscv64-unknown-elf-g++-11.1.0 regression suite reports the following
failures for
$ make check-gcc-c++ RUNTESTFLAGS='dg.exp=Wstringop-overflow-6.C'
```
FAIL: g++.dg/warn/Wstringop-overflow-6.C -std=gnu++14 (test for excess errors)
FAIL: g++.dg/warn/Wstringop-overflow-6.C -std=gnu++17 (test for excess errors)
FAIL: g++.dg/warn/Wstringop-overflow-6.C -std=gnu++2a (test for excess errors)
UNSUPPORTED: g++.dg/warn/Wstringop-overflow-6.C -std=gnu++98
```
The "excess errors" being
```
output is In file included from /home/maxim/prj/riscv-upstream/install/riscv64-unknown-elf/include/wchar.h:6,
from /home/maxim/prj/riscv-upstream/build/gcc-stage2/riscv64-unknown-elf/libstdc++-v3/include/cwchar:44,
from /home/maxim/prj/riscv-upstream/build/gcc-stage2/riscv64-unknown-elf/libstdc++-v3/include/bits/postypes.h:40,
from /home/maxim/prj/riscv-upstream/build/gcc-stage2/riscv64-unknown-elf/libstdc++-v3/include/iosfwd:40,
from /home/maxim/prj/riscv-upstream/build/gcc-stage2/riscv64-unknown-elf/libstdc++-v3/include/ios:38,
from /home/maxim/prj/riscv-upstream/build/gcc-stage2/riscv64-unknown-elf/libstdc++-v3/include/ostream:38,
from /home/maxim/prj/riscv-upstream/build/gcc-stage2/riscv64-unknown-elf/libstdc++-v3/include/iostream:39,
from /home/maxim/prj/riscv-upstream/gcc-11.1.0/gcc/testsuite/g++.dg/warn/Wstringop-overflow-6.C:6:
/home/maxim/prj/riscv-upstream/install/riscv64-unknown-elf/include/sys/reent.h:685:11: warning: unnecessary parentheses in declaration of '_sig_func' [-Wparentheses]
```
cc Aldy Hernandez <aldyh@redhat.com> and Andrew MacLeod <amacleod@redhat.com>,
they are author of new VRP analysis for GCC, just to make sure I didn't
mis-understanding or mis-interpreting anything on GCC site.
GCC 11 have better value range analysis, that give GCC more confidence
to perform more aggressive optimization, but it cause scalbn/scalbnf get
wrong result.
Using scalbn to demostrate what happened on GCC 11, see comments with VRP
prefix:
```c
double scalbn (double x, int n)
{
/* VRP RESULT: n = [-INF, +INF] */
__int32_t k,hx,lx;
...
k = (hx&0x7ff00000)>>20;
/* VRP RESULT: k = [0, 2047] */
if (k==0) {
/* VRP RESULT: k = 0 */
...
k = ((hx&0x7ff00000)>>20) - 54;
if (n< -50000) return tiny*x; /*underflow*/
/* VRP RESULT: k = -54 */
}
/* VRP RESULT: k = [-54, 2047] */
if (k==0x7ff) return x+x; /* NaN or Inf */
/* VRP RESULT: k = [-54, 2046] */
k = k+n;
if (k > 0x7fe) return huge*copysign(huge,x); /* overflow */
/* VRP RESULT: k = [-INF, 2046] */
/* VRP RESULT: n = [-INF, 2100],
because k + n <= 0x7fe is false, so:
1. -INF < [-54, 2046] + n <= 0x7fe(2046) < INF
2. -INF < [-54, 2046] + n <= 2046 < INF
3. -INF < n <= 2046 - [-54, 2046] < INF
4. -INF < n <= [0, 2100] < INF
5. n = [-INF, 2100] */
if (k > 0) /* normal result */
{SET_HIGH_WORD(x,(hx&0x800fffff)|(k<<20)); return x;}
if (k <= -54) {
/* VRP OPT: Evaluate n > 50000 as true...*/
if (n > 50000) /* in case integer overflow in n+k */
return huge*copysign(huge,x); /*overflow*/
else return tiny*copysign(tiny,x); /*underflow*/
}
k += 54; /* subnormal result */
SET_HIGH_WORD(x,(hx&0x800fffff)|(k<<20));
return x*twom54;
}
```
However give the input n = INT32_MAX, k = k+n will overflow, and then we
expect got `huge*copysign(huge,x)`, but new VRP optimization think
`n > 50000` is never be true, so optimize that into `tiny*copysign(tiny,x)`.
so the solution here is to moving the overflow handle logic before `k = k + n`.
- GCC will set __FLT_EVAL_METHOD__ to 16 if __fp16 supported, e.g.
cortex-a55/aarch64.
- $ aarch64-unknown-elf-gcc -v 2>&1 |grep version
gcc version 9.2.0 (GCC)
- $ aarch64-unknown-elf-gcc -E -dM -mcpu=cortex-a55 - < /dev/null |grep FLT_EVAL_METHOD
#define __FLT_EVAL_METHOD__ 16
#define __FLT_EVAL_METHOD_TS_18661_3__ 16
#define __FLT_EVAL_METHOD_C99__ 16
- The behavior of __FLT_EVAL_METHOD__ == 16 is same as
__FLT_EVAL_METHOD__ == 0 except for float16_t, but newlib didn't
support float16_t.
ISO/IEC TS 18661-3:
http://www.open-std.org/jtc1/sc22/wg14/www/docs/n2405.pdf
V2 Changes:
- List Howland, Craig D as co-author since he provide the draft of comment
in math.h.
Co-authored-by: "Howland, Craig D" <howland@LGSInnovations.com>
The C standard says that errno may acquire the value ERANGE if the
result from strtod underflows. According to IEEE 754, underflow occurs
whenever the value cannot be represented in normalized form.
Newlib is inconsistent in this, setting errno to ERANGE only if the
value underflows to zero, but not for denorm values, and never for hex
format floats.
This patch attempts to consistently set errno to ERANGE for all
'underflow' conditions, which is to say all values which are not
exactly zero and which cannot be represented in normalized form.
This matches glibc behavior, as well as the Linux, Mac OS X, OpenBSD,
FreeBSD and SunOS strtod man pages.
Signed-off-by: Keith Packard <keithp@keithp.com>
The scanf code was skipping the '0' after the 'x' causing the
resulting buffer to contain an invalid number when passed to strtod.
Signed-off-by: Keith Packard <keithp@keithp.com>
- compiler is sometimes optimizing out the rounding check in
e_sqrt.c and ef_sqrt.c which uses two constants to create
an inexact operation
- there is a similar constant operation in s_tanh.c/sf_tanh.c
- make the one and tiny constants volatile to stop this
Newlib for aarch64 uses libgloss for the backend. One common libgloss
implementation is the 'rdimon' implementation, which uses the Arm
Semihosting protocol. In order to support a remote host that runs on
Windows we need to know whether a file is to be opened in binary or
text mode. That means that we need to preserve this information via
O_BINARY until we know what the libgloss binding will be.
This patch simply copies the arm implementation from sys/arm/sys and
puts it in machine/aarch64/sys, because we don't have a 'sys' subtree
on aarch64.
The only reason why it is tough for us to use nano malloc
is because of the small shortcoming where nano_malloc()
splits a bigger chunk from the free list into two pieces
while handing back the second one (the tail) to the user.
This is error prone and especially bad for smaller heaps,
where nano malloc is supposed to be superior. The normal
malloc doesn't have this issue and we need to use it even
though it costs us ~2k bytes compared to nano-malloc.
The problem arise especially after giving back _every_
malloced memory to the heap and then starting to exercise
the heap again by allocating something small. This small
item might split the whole heap in two equally big parts
depending on how the heap has been exercised before.
I have uploaded the smallest possible application
(only tested on ST and Nordic devices) to show the issue
while the real customer applications are far more complicated:
https://drive.google.com/file/d/1kfSC2KOm3Os3mI7EBd-U0j63qVs8xMbt/view?usp=sharing
The application works like the following pseudo code,
where we assume a heap of 100 bytes
(I haven't taken padding and other nitty and gritty
details into account. Everything to simplify understanding):
void *ptr = malloc(52); // We get 52 bytes and we have
// 48 bytes to use.
free(ptr); // Hand back the 52 bytes to nano_malloc
// This is the magic line that shows the issue of
// nano_malloc
ptr = malloc(1); // Nano malloc will split the 52 bytes
// in the free list and hand you a pointer
// somewhere in the
// middle of the heap.
ptr2 = malloc(52); // Out of memory...
I have done a fix which hands back the first part of the
splitted chunk. Once this is fixed we obviously
have the 1 byte placed in position 0 of the heap instead
of somewhere in the middle.
However, this won't let us malloc 52 new bytes even though
we potentially have 99 bytes left to use in the heap. The
reason is that when we try to do the allocation,
nano-malloc looks into the free list and sees a 51 byte
chunk to be used.
This is not big enough so nano-malloc decides to call
sbrk for _another_ 52 bytes which is not possible since
there is only 48 bytes left to ask for.
The solution for this problem is to check if the last
item in the free list is adjacent to sbrk(0). If it is,
as it is in this case, we can just ask sbrk for the
remainder of what is needed. In this case 1 byte.
NB! I have only tested the solution on our ST device.
Use the more official fesetenv(FE_DFL_ENV) from _dll_crt0, thus
allowing to drop the _feinitialise declaration from fenv.h.
Provide a no-op _feinitialise in Cygwin as exportable symbol for really
old applications when _feinitialise was called from mainCRTStartup in
crt0.o.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Drop the Cygwin-specific fenv.cc and fenv.h file and use the equivalent
newlib functionality now, so we have at least one example of a user for
this new mechanism.
fenv.c: allow _feinitialise to be called from Cygwin startup code
fenv.h: add declarations for fegetprec and fesetprec for Cygwin only.
Fix a comment.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Add a directory libc/machine/shared_x86 to share header files
between ix86 and x86_64 architectures.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
So far the build mechanism in newlib only allowed to either define
machine-specific headers, or headers shared between all machines.
In some cases, architectures are sufficiently alike to share header
files between them, but not with other architectures. A good example
is ix86 vs. x86_64, which share certain traits with each other, but
not with other architectures.
Introduce a new configure variable called "shared_machine_dir". This
dir can then be used for headers shared between architectures.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Without this, for a bare-iron/simulator target such as cris-elf,
you'll see, at newlib build time:
/x/gccobj/./gcc/xgcc -B/x/gccobj/./gcc/ <many options elided> -c -o lib_a-aligned_alloc.o \
`test -f 'aligned_alloc.c' || echo '/y/newlib/libc/stdlib/'`aligned_alloc.c
/y/newlib/libc/stdlib/aligned_alloc.c: In function 'aligned_alloc':
/y/newlib/libc/stdlib/aligned_alloc.c:35:10: warning: implicit declaration of function \
'_memalign_r' [-Wimplicit-function-declaration]
35 | return _memalign_r (_REENT, align, size);
| ^~~~~~~~~~~
The revert-part of the revert-and-fix commit, b99887c428 a.k.a.
"Revert previous change to sys/stat.h and fix cris libgloss",
apparently intending to revert f75aa67851 a.k.a. "Fix regression in
cris-elf caused by sys/stat.h change" and fix it in another way,
wasn't complete. Although the fix-part added the prerequisite "#undef
st_atime" (et al) to gensyscalls, the revert-part didn't revert the
"&& !defined(__cris__)" in sys/stat.h, stopping st_atime (et al) from
being defined.
The effect of the unreverted change is that accessing the struct stat
compatibility member names "st_atime" (et al) as in "struct stat
mystat; mystat.st_atime;" yields errors, observable for example when
building libgfortran in gcc:
/x/gcc/libgfortran/intrinsics/stat.c:114:42: error: 'struct stat' has \
no member named 'st_atime'; did you mean 'st_atim'?
114 | sarray->base_addr[8 * stride] = sb.st_atime;
| ^~~~~~~~
| st_atim
(etc.)
Trivially fixed by completing the reversion, removing the "&&
!defined(__cris__)" in sys/stat.h.
Beware: the net effect of the earlier related change to struct stat in
sys/stat.h, leading up to the fix, *does* change its definition as a
type. Thankfully, replacing members like "time_t st_atime; long
st_spare1;" by "struct timespec st_atim;", ditto st_mtim and st_ctim,
is layout-compatible. To wit, that change is "binary compatible".
Incidentally, related to the simulator / Linux ABI, there's a
transitional stage (see gensyscalls), reloading between "struct stat"
(sys/stat.h) and "struct new_stat" (kernel/simulator) as necessary.
Tested by a cris-elf gcc build (including libgfortran).
This Patch removes Soft Float code from MIPS.
Instead It adds the soft float code from RISCV
The code came from FreeBSD and assumes the FreeBSD softfp
implementation not the one with GCC. That was an overlooked and
fixed in the other fenv code already.
Signed-off-by: Eshan Dhawan <eshandhawan51@gmail.com>
Change the prototypes to be in line with POSIX/glibc. This may fix
issues with new warnings produced by GCC 11.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Add the POSIX header file <poll.h> which is used by the GCC 11 Ada
runtime support.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
The Cortex-R52 processor is an Armv8-R processor with a NEON unit. This
fix prevents conflicting architecture profiles A/R errors issued by the
linker.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
This patch fixes the error found by Paul Zimmermann (see
https://homepages.loria.fr/PZimmermann/papers/#accuracy) regarding x
close to 1 and rather large y (specifically he found the case
powf(0x1.ffffeep-1,-0x1.000002p+27) which returns +Inf instead of the
correct value). We found 2 more values for x which show the same faulty
behaviour, and all 3 are fixed with this patch. We have tested all
combinations for x in [+1.fffdfp-1, +1.00020p+0] and y in
[-1.000007p+27, -1.000002p+27] and [1.000002p+27,1.000007p+27].
The overflow check in mEMALIGn erroneously checks for INT_MAX,
albeit the input parameter is size_t. Fix this to check for
__SIZE_MAX__ instead. Also, it misses to check the req against
adding the alignment before calling mALLOc.
While at it, add out-of-bounds checks to pvALLOc, nano_memalign,
nano_valloc, and Cygwin's (unused) dlpvalloc.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
to rb parent pointers. Define RB_SWAP_CHILD to replace the child of a parent
with its twin, and use it in 4 places. Use RB_SET in rb_link_node to remove the
only linuxkpi reference to color, and then drop color- and parent-related
definitions that are defined and used only in rbtree.h.
This is intended to be entirely cosmetic, with no impact on program
behavior, and leave RB_PARENT and RB_SET_PARENT as the only ways to
read and write rb parent pointers.
Reviewed by: markj, kib
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D25264
the sibling of the root of the too-short tree is black and at least one of the
children of that sibling is red, either one or two rotations finish the
rebalancing. In the case when both of the children are red, the current
implementation uses two rotations where only one is necessary. This change
removes that extra rotation, and in that case also removes a needless
black-to-red-to-black recoloring.
Reviewed by: markj
Differential Revision: https://reviews.freebsd.org/D25335
without using their interfaces, making them break when the representation
changes. Revert changes that eliminated the color field from rb-trees, leaving
everything as it was before.
Reviewed by: markj
Differential Revision: https://reviews.freebsd.org/D25250
a bit of DIAGNOSTIC code that depends on it.
Reported by: rpokala, mjguzik
Reviewed by: markj
Differential Revision: https://reviews.freebsd.org/D25204
field. Set the least significant bit in the pointer to the node from its parent
to indicate that the node is red. Have the tree rotation macros leave the
old-parent/new-child node red and the new-parent/old-child node black.
This change makes RB_LEFT and RB_RIGHT no longer assignable, and
RB_COLOR no longer defined. Any code that modifies the tree or
examines a node color would have to be modified after this change.
Reviewed by: markj
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D25105
where the pointer checked is provably never null. Restructure the surrounding
code just enough to make the non-nullness obvious.
Reviewed by: markj
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D25089
child is red or child is null. In the first case, RB_REMOVE_COLOR just changes
the child to black and returns. With this change, RB_REMOVE handles that case,
and drops the child argument to RB_REMOVE_COLOR, since that value is always
null.
RB_REMOVE_COLOR is changed to remove a couple of unneeded tests, and
to eliminate some deep indentation.
RB_ISRED is defined to combine a null check with a test for redness,
to replace that combination in several places.
Reviewed by: markj
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D25032
search to find the node to replace the one being removed, restructure to first
remove the replacement node and correct the parent pointers around it, and then
let the all-cases code at the end deal with the parent of the deleted node,
making it point to the replacement node. This removes one or two conditional
branches.
Reviewed by: markj
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D24845
macros so that is invoked at the root of every subtree that changes in an
insert or delete, and only once, and ordered from the bottom of the tree to the
top. For intel_gas.c, the only user of RB_AUGMENT I can find, change the
augmenting routine so that it does not climb from entry to tree root on every
call, and remove a 'tree correcting' function that can be supplanted by proper
tree augmentation.
Reviewed by: kib
Tested by: pho
Differential Revision: https://reviews.freebsd.org/D23189
removing a node and reinserting it back with an updated key.
This is one of dependencies for the upcoming stats(3) code.
Reviewed by: cem
Obtained from: Netflix
MFC after: 2 weeks
Sponsored by: Klara Inc, Netflix
Differential Revision: https://reviews.freebsd.org/D21786
For configurations without x2APIC support (guests, older hardware), the global
LAPIC MMIO mapping will trigger false-positive KCSan reports as it will appear
that multiple CPUs are concurrently reading and writing the same address.
This isn't actually true, as the underlying physical access will be performed
on the local CPU's APIC. Additionally, because LAPIC access can happen during
event timer configuration, the resulting KCSan printf can produce a panic due
to attempted recursion on event timer resources.
Add a __nosanitizethread preprocessor define to prevent the compiler from
inserting TSan hooks, and apply it to the x86 LAPIC accessors.
PR: 249149
Reported by: gbe
Reviewed by: andrew, kib
Tested by: gbe
Differential Revision: https://reviews.freebsd.org/D26354
This is an unfortunate instance where the __has_attribute check does
not function usefully. Gcc does have the attribute, but for gcc it only
applies to functions, not variables, and trying to apply it to a
variable generates Wattribute. So far we only apply the attribute to
variables. Only enable the attribute for clang, for now.
Reviewed by: Anton Rang <rang at acm.org>
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D22875
The assumptions of linker_set don't play nicely with
AddressSanitizer. AddressSanitizer adds a 'redzone' of zeros around
globals (including those in named sections), whereas linker_set
assumes they are all packed consecutively like a pointer array. So:
let's annotate linker_set so that AddressSanitizer ignores it.
Submitted by: Matthew Bryan <matthew.bryan@isilon.com>
Reviewed by: kib, rang_acm.org
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D22239
The <sys/cdefs.h> and <stdatomic.h> headers already included support for
C11 atomics via intrinsincs in modern versions of GCC, but these versions
tried to "hide" atomic variables inside a wrapper structure. This wrapper
is not compatible with GCC's internal <stdatomic.h> header, so that if
GCC's <stdatomic.h> was used together with <sys/cdefs.h>, use of C11
atomics would fail to compile. Fix this by not hiding atomic variables
in a structure for modern versions of GCC. The headers already avoid
using a wrapper structure on clang.
Note that this wrapper was only used if C11 was not enabled (e.g.
via -std=c99), so this also fixes compile failures if a modern version
of GCC was used with -std=c11 but with FreeBSD's <stdatomic.h> instead
of GCC's <stdatomic.h> and this change fixes that case as well.
Reported by: Mark Millard
Reviewed by: kib
Differential Revision: https://reviews.freebsd.org/D16585
As discussed in GCC bug 97088
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97088), parameters in
prototypes of library functions should use reserved names, or no name
at all.
This patch moves the internal struct __tzrule_struct to its own
internal header sys/_tz_structs.h. This is included from newlib's
time code as well as from Cygwin's localtime wrapper.
Signed-off-by: Torbjörn SVENSSON <torbjorn.svensson@st.com>
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
As discussed in GCC bug 97088
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97088), parameters in
prototypes of library functions should use reserved names, or no name
at all.
This patch removes the 'ptr' parameter name from
wint_t _getwchar_r (struct _reent *);
wint_t _getwchar_unlocked_r (struct _reent *);
to avoid possible clashes with user code in case someone uses
before including Newlib's wchar.h (or uses some other conflicting
definition)
Signed-off-by: Torbjörn SVENSSON <torbjorn.svensson@st.com>
As discussed in GCC bug 97088
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97088), parameters in
prototypes of library functions should use reserved names, or no name
at all.
This patch removes the 'j' parameter name from
extern intmax_t imaxabs(intmax_t);
to avoid possible clashes with user code in case someone uses
before including Newlib's inttypes.h (or uses some other conflicting
definition)
Signed-off-by: Torbjörn SVENSSON <torbjorn.svensson@st.com>
As discussed in GCC bug 97088
(https://gcc.gnu.org/bugzilla/show_bug.cgi?id=97088), parameters in
prototypes of library functions should use reserved names, or no name
at all.
This patch removes the 'x' parameter name from
extern int __isinff (float);
extern int __isinfd (double);
extern int __isnanf (float);
extern int __isnand (double);
extern int __fpclassifyf (float);
extern int __fpclassifyd (double);
extern int __signbitf (float);
extern int __signbitd (double);
to avoid possible clashes with user code in case someone uses
before including Newlib's math.h (or uses some other conflicting
definition)
The current gamma, gamma_r, gammaf and gammaf_r functions return
|gamma(x)| instead of ln(|gamma(x)|) due to a change made back in 2002
to the __ieee754_gamma_r implementation. This patch fixes that, making
all of these functions map too their lgamma equivalents.
To fix the underlying bug, the __ieee754_gamma functions have been
changed to return gamma(x), removing the _r variants as those are no
longer necessary. Their names have been changed to __ieee754_tgamma to
avoid potential confusion from users.
Now that the __ieee754_tgamma functions return the correctly signed
value, the tgamma functions have been modified to use them.
libm.a now exposes the following gamma functions:
ln(|gamma(x)|):
__ieee754_lgamma_r
__ieee754_lgammaf_r
lgamma
lgamma_r
gamma
gamma_r
lgammaf
lgammaf_r
gammaf
gammaf_r
lgammal (on machines where long double is double)
gamma(x):
__ieee754_tgamma
__ieee754_tgammaf
tgamma
tgammaf
tgammal (on machines where long double is double)
Additional aliases for any of the above functions can be added if
necessary; in particular, I'm not sure if we need to include
__ieee754_gamma*_r functions (which would return ln(|(gamma(x)|).
Signed-off-by: Keith Packard <keithp@keithp.com>
----
v2:
Switch commit message to ASCII
For RISC-V targets without hardware FMA support, include the
common fma implementation to provide that API.
Signed-off-by: Keith Packard <keithp@keithp.com>
Like ARM, some RISC-V implementations have hardware sqrt. Support for
that can be detected at compile time, which the code did. However, the
filenames were incorrect so that both the risc-v specific and general
code were getting included in the resulting library.
Fix this by following the ARM model and #include'ing the general code
when the architecture-specific support is not available.
Signed-off-by: Keith Packard <keithp@keithp.com>
The MSP430 target supports both 16-bit and 20-bit size_t and intptr_t.
Some implicit casts in Newlib expect these types to be
"long", (a 32-bit type on MSP430) which causes warnings during
compilation such as:
"cast from pointer to integer of different size"
This is required to avoid colliding with files built from libm/common
that would end up with the same object name.
When libm.a was constructed from the individual sub-libraries, the
contents of the libm/common files would be replaced by that from
libm/machine/arm with the same name.
Signed-off-by: Keith Packard <keithp@keithp.com>
commit 588a5e1dde added a non-reentrant
call to nano_malloc which causes a build failure if INTERNAL_NEWLIB is
defined.
Here is a snippet of the error:
In file included from .../newlib/newlib/libc/stdlib/nano-mallocr.c:38:
.../newlib/newlib/libc/include/malloc.h:42:25: note: expected 'struct _reent *' but argument is of type 'ptrdiff_t' {aka 'int'}
42 | extern void *_malloc_r (struct _reent *, size_t);
| ^~~~~~~~~~~~~~~
.../newlib/newlib/libc/stdlib/nano-mallocr.c:67:22: error: too few arguments to function '_malloc_r'
67 | #define nano_malloc _malloc_r
| ^~~~~~~~~
.../newlib/newlib/libc/stdlib/nano-mallocr.c:456:11: note: in expansion of macro 'nano_malloc'
456 | mem = nano_malloc(bytes);
| ^~~~~~~~~~~
In file included from .../newlib/newlib/libc/stdlib/nano-mallocr.c:38:
.../newlib/newlib/libc/include/malloc.h:42:14: note: declared here
42 | extern void *_malloc_r (struct _reent *, size_t);
| ^~~~~~~~~
.../newlib/newlib/libc/stdlib/nano-mallocr.c:43: warning: "assert" redefined
43 | #define assert(x) ((void)0)
|
This patch adds a missing RCALL to the args when calling nano_malloc
from nano_calloc, so that if the call is reentrant, reent_ptr is passed
as the first argument.
The variable `bytes` (also added in 588a5e1d) has been changed from a
`ptrdiff_t` to `malloc_size_t` as it does not need to be signed. It is
used to store the product of two unsigned malloc_size_t variables and
then iff there was no overflow is it passed to malloc and memset which
both expect size_t which is unsigned.
Signed-off-by: Craig Blackmore <craig.blackmore@embecosm.com>
This built-in function (available in both gcc and clang) is more
efficient and generates shorter code than open-coding the test.
Signed-off-by: Keith Packard <keithp@keithp.com>
When HAVE_FAST_FMAF is set, use the vfma.f32 instruction, when
HAVE_FAST_FMA is set, use the vfma.f64 instruction.
Usually the compiler built-ins will already have inlined these
instructions, but provide these symbols for cases where that doesn't
work instead of falling back to the (inaccurate) common code versions.
Signed-off-by: Keith Packard <keithp@keithp.com>
Anything with fast FMA is assumed to have fast FMAF, along with
32-bit arms that advertise 32-bit FP support and __ARM_FEATURE_FMA
Signed-off-by: Keith Packard <keithp@keithp.com>
32-bit ARM processors with HW float (but not HW double) may define
__ARM_FEATURE_FMA, but that only means they have fast FMA for 32-bit
floats.
Signed-off-by: Keith Packard <keithp@keithp.com>
It was calling __math_uflow(0) instead of __math_uflowf(0), which
resulted in no exception being set on machines with exception support
for float but not double.
Signed-off-by: Keith Packard <keithp@keithp.com>
This removes the run-time configuration of errno support present in
portions of the math library and unifies all of the compile-time errno
configuration under a single parameter so that the whole library
is consistent.
The run-time support provided by _LIB_VERSION is no longer present in
the public API, although it is still used internally to disable errno
setting in some functions. Now that it is a constant, the compiler should
remove that code when errno is not supported.
This removes s_lib_ver.c as _LIB_VERSION is no longer variable.
Signed-off-by: Keith Packard <keithp@keithp.com>
The __ieee754 functions already return the right value in exception
cases, so don't modify those. Setting the library to _POSIX_/_IEEE_
mode now only affects whether errno is modified.
Signed-off-by: Keith Packard <keithp@keithp.com>
The y0, y1 and yn functions need separate conditions when x is zero as
that returns ERANGE instead of EDOM.
Also stop adjusting the return value from the __ieee754_y* functions
as that is already correct and we were just breaking it.
Signed-off-by: Keith Packard <keithp@keithp.com>
math_errhandling is specified to contain two bits of information:
1. MATH_ERRNO -- Set when the library sets errno
2. MATH_ERREXCEPT -- Set when math operations report exceptions
MATH_ERRNO should match whether the original math code is compiled in
_IEEE_LIBM mode and the new math code has WANT_ERRNO == 1.
MATH_ERREXCEPT should match whether the underlying hardware has
exception support. This patch adds configurations of this value for
RISC-V, ARM, Aarch64, x86 and x86_64 when using HW float.
Signed-off-by: Keith Packard <keithp@keithp.com>
_IEEE_LIBM is the configuration value which controls whether the
original libm functions modify errno. Use that in the new math code as
well so that the resulting library is internally consistent.
Signed-off-by: Keith Packard <keithp@keithp.com>
C compilers may fold const values at compile time, so expressions
which try to elicit underflow/overflow by performing simple
arithemetic on suitable values will not generate the required
exceptions.
Work around this by replacing code which does these arithmetic
operations with calls to the existing __math_xflow functions that are
designed to do this correctly.
Signed-off-by: Keith Packard <keithp@keithp.com>
----
v2:
libm/math: Pass sign to __math_xflow instead of muliplying result
Compiling
#include <sys/select.h>
void f(int X)
{
fd_set set;
FD_ZERO(&set);
FD_SET(X,&set);
FD_CLR(X+1,&set);
(void)FD_ISSET(X+2,&set);
}
results in plenty of gcc warnings when compiled with
-Wconversion -Wsign-conversion:
fds.c:7:2: warning: conversion to ‘long unsigned int’ from ‘int’ may
FD_SET(X,&set);
^~~~~~
[...]
The unsigned NFDBITS macro combined with the signed 1L constant
are causing lots of implicit signed/unsigned type conversions.
Fix this by updating the FD_* macro code to the latest from FreeBSD
and adding an (int) cast to _NFDBITS.
As a side-effect, this fixes the visibility of NFDBITS and
fds_bits (only if __BSD_VISIBLE).
This also eliminates the old, outdated fd_set workaround.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
Newlib's posix_spawn has been taken from FreeBSD. The code relies on
BSD-specific behaviour of vfork, namely the fact that vfork blocks
the parent until the child exits or calls execve as well as the fact
that the child shares parent memory in non-COW mode.
This behaviour can't be emulated by Cygwin. Cygwin's vfork is
equivalent to fork. This is POSIX-compliant, but it's lacking BSD's
vfork ingrained synchronization of the parent to wait for the child
calling execve, or the chance to just write a variable and the parent
will see the result.
So this requires a Cygwin-specific solution. The core function of
posix_spawn, called do_posix_spawn is now implemented twice, once using
the BSD method, and once for Cygwin using Windows synchronization under
the hood waiting for the child to call execve and signalling errors
upstream. The Windows specifics are hidden inside Cygwin, so newlib
only calls internal Cygwin functions.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
ld: libm.a(lib_a-fesetenv.o): in function `fesetenv':
newlib/libm/machine/arm/fesetenv.c:38: undefined reference to `vmsr_fpscr'
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Use the already existing stub files if possible. These files are
necessary to override the stub implementation with the machine-specific
implementation through the build system.
Reviewed-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Signed-off-by: Eshan dhawan <eshandhawan51@gmail.com>
The previous fenv support for ARM used the soft-float implementation of
FreeBSD. Newlib uses the one from libgcc by default. They are not
compatible. Having an GCC incompatible soft-float fenv support in
Newlib makes no sense. A long-term solution could be to provide a
libgcc compatible soft-float support. This likely requires changes in
the GCC configuration. For now, provide a stub implementation for
soft-float multilibs similar to RISC-V.
Move implementation to one file and delete now unused files. Hide
implementation details. Remove function parameter names from header
file to avoid name conflicts.
Provide VFP support if __SOFTFP__ is not defined like glibc.
Reviewed-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Signed-off-by: Eshan dhawan <eshandhawan51@gmail.com>
This patch fixes a bug in RISC-V's memcpy implementation where an
integer wraparound occurs when src + size < 8 * sizeof(long), causing
the word-sized copy loop to be incorrectly entered.
Signed-off-by: Chih-Mao Chen <cmchen@andestech.com>
If __HAVE_LOCALE_INFO__ is not defined, then the locale in the
locale-specific ctype functions is ignored. In the previous
implementation this resulted in compiler warnings. For example:
int main()
{
locale_t locale;
locale = duplocale(uselocale((locale_t)0));
isspace_l('x', locale);
return 0;
}
gcc -Wall main.c
main.c: In function 'main':
main.c:6:11: warning: variable 'locale' set but not used [-Wunused-but-set-variable]
6 | locale_t locale;
| ^~~~~~
_ICONV_CONVERTER -> ICONV_FROM_ENCODING. It's not perfect, as the
library can support different from/to encodings now, but at least in
the default configurations the tests now work.
Signed-off-by: Keith Packard <keithp@keithp.com>
This caused the strnstr to walk off the end of the alias array and
fetch invalid data. Instead of attempting to update 'len', just
re-compute it based on the table end pointer that is already known.
Signed-off-by: Keith Packard <keithp@keithp.com>
The pointer value for the iconv alias data never changes, so get rid
of the pointer and make it an array instead.
Signed-off-by: Keith Packard <keithp@keithp.com>
Fix the code checking for character set loading failure so that
it checks the return value from the init function.
Signed-off-by: Keith Packard <keithp@keithp.com>
The original implementation had multiple issues:
- Only worked when posix_memalign was available (Linux, RTEMS).
- Violated C11 link namespace rules by calling posix_memalign.
- Failed to set errno on error.
These can be fixed by essentially using the same implementation
for aligned_alloc as for memalign, i.e. simply calling _memalign_r
(which is always available and a "more reserved name" although
technically still not in the reserved link namespace, at least
code written in c cannot define a colliding symbol, newlib has
plenty such namespace issues so this is fine).
It is not clear what the right policy is when MALLOC_PROVIDED is set,
currently that does not cover aligned_alloc so it is kept that way.
Tested on aarch64-none-elf
The IEEE spec for pow only has special case for x**0 and 1**y when x/y
are quiet NaN. For signaling NaN, the general case applies and these functions
should signal the invalid exception and return a quiet NaN.
Signed-off-by: Keith Packard <keithp@keithp.com>
These functions shared a pattern of re-converting the argument to bits
when returning +/-0. Skip that as the initial conversion still has the
sign bit.
Signed-off-by: Keith Packard <keithp@keithp.com>
Recent GCC appears to elide multiplication by 1, which causes snan
parameters to be returned unchanged through *iptr. Use the existing
conversion of snan to qnan to also set the correct result in *iptr
instead.
Signed-off-by: Keith Packard <keithp@keithp.com>
This fix comes from glibc, from files which originated from
the same place as the newlib files. Those files in glibc carry
the same license as the newlib files.
Bug 14155 is spurious underflow exceptions from Bessel functions for
large arguments. (The correct results for large x are roughly
constant * sin or cos (x + constant) / sqrt (x), so no underflow
exceptions should occur based on the final result.)
There are various places underflows may occur in the intermediate
calculations that cause the failures listed in that bug. This patch
fixes problems for the double version where underflows occur in
calculating the intermediate functions P and Q (in particular, x**-12
gets computed while calculating Q). Appropriate approximations are
used for P and Q for arguments at least 0x1p28 and above to avoid the
underflows.
For sufficiently large x - 0x1p129 and above - the code already has a
cut-off to avoid calculating P and Q at all, which means the
approximations -0.125 / x and 0.375 / x can't themselves cause
underflows calculating Q. This cut-off is heuristically reasonable
for the point beyond which Q can be neglected (based on expecting
around 0x1p-64 to be the least absolute value of sin or cos for large
arguments representable in double).
The float versions use a cut-off 0x1p17, which is less heuristically
justifiable but should still only affect values near zeroes of the
Bessel functions where these implementations are intrinsically
inaccurate anyway (bugs 14469-14472), and should serve to avoid
underflows (the float underflow for jn in bug 14155 probably comes
from the recurrence to compute jn). ldbl-96 uses 0x1p129, which may
not really be enough heuristically (0x1p143 or so might be safer - 143
= 64 + 79, number of mantissa bits plus total number of significant
bits in representation) but again should avoid underflows and only
affect values where the code is substantially inaccurate anyway.
ldbl-128 and ldbl-128ibm share a completely different implementation
with no such cut-off, which I propose to fix separately.
Signed-off-by: Keith Packard <keithp@keithp.com>
Add the missing mask for the decomposition of hi+lo which caused some
errors of 1-2 ULP.
This change is taken over from FreeBSD:
95436ce20d
Additionally I've removed some variable assignments which were never
read before being overwritten again in the next 2 lines.
This fix for k_tan.c is a copy from fdlibm version 5.3 (see also
http://www.netlib.org/fdlibm/readme), adjusted to use the macros
available in newlib (SET_LOW_WORD).
This fix reduces the ULP error of the value shown in the fdlibm readme
(tan(1.7765241907548024E+269)) to 0.45 (thereby reducing the error by
1).
This issue only happens for large numbers that get reduced by the range
reduction to a value smaller in magnitude than 2^-28, that is also
reduced an uneven number of times. This seems rather unlikely given that
one ULP is (much) larger than 2^-28 for the values that may cause an
issue. Although given the sheer number of values a double can
represent, it is still possible that there are more affected values,
finding them however will be quite hard, if not impossible.
We also took a look at how another library (libm in FreeBSD) handles the
issue: In FreeBSD the complete if branch which checks for values smaller
than 2^-28 (or rather 2^-27, another change done by FreeBSD) is moved
out of the kernel function and into the external function. This means
that the value that gets checked for this condition is the unreduced
value. Therefore the input value which caused a problem in the
fdlibm/newlib kernel tan will run through the full polynomial, including
the careful calculation of -1/(x+r). So the difference is really whether
r or y is used. r = y + p with p being the result of the polynomial with
1/3*x^3 being the largest (and magnitude defining) value. With x being
<2^-27 we therefore know that p is smaller than y (y has to be at least
the size of the value of x last mantissa bit divided by 2, which is at
least x*2^-51 for doubles) by enough to warrant saying that r ~ y. So
we can conclude that the general implementation of this special case is
the same, FreeBSD simply has a different philosophy on when to handle
especially small numbers.
Make line 47 in sf_trunc.c reachable. While converting the double
precision function trunc to the single precision version truncf an error
was introduced into the special case. This special case is meant to
catch both NaNs and infinities, however qNaNs and infinities work just
fine with the simple return of x (line 51). The only error occurs for
sNaNs where the same sNaN is returned and no invalid exception is
raised.
The comparison c == FP_INFINITE causes the function to return +inf as it
expects x = +inf to always be larger than y. This shortcut causes
several issues as it also returns +inf for the following cases:
- fdim(+inf, +inf), expected (as per C99): +0.0
- fdim(-inf, any non NaN), expected: +0.0
I don't see a reason to keep the comparison as all the infinity cases
return the correct result using just the ternary operation.
While testing the exp function we noticed some errors at the specified
magnitude. Within this range the exp function returns the input value +1
as an output. We chose to run a test of 1m exponentially spaced values
in the ranges [-2^-27,-2^-32] and [2^-32,2^-27] which showed 7603 and
3912 results with an error of >=0.5 ULP (compared with MPFR in 128 bit)
with the highest being 0.56 ULP and 0.53 ULP.
It's easy to fix by changing the magnitude at which the input value +1
is returned from <2^-28 to <2^-32 and using the polynomial instead. This
reduces the number of results with an error of >=0.5 ULP to 485 and 479
in above tests, all of which are exactly 0.5 ULP.
As we were already checking on exp we also took a look at expf. For expf
the magnitude where the input value +1 is returned can be increased from
<2^-28 to <2^-23 without accuracy loss for a slight performance
improvement. To ensure this was the correct value we tested all values
in the ranges [-2^-17,-2^-28] and [2^-28,2^-17] (~92.3m values each).
The single-precision trigonometric functions show rather high errors in
specific ranges starting at about 30000 radians. For example the sinf
procedure produces an error of 7626.55 ULP with the input
5.195880078125e+04 (0x474AF6CD) (compared with MPFR in 128bit
precision). For the test we used 100k values evenly spaced in the range
of [30k, 70k]. The issues are periodic at higher ranges.
This error was introduced when the double precision range reduction was
first converted to float. The shift by 8 bits always returns 0 as iq is
never higher than 255.
The fix reduces the error of the example above to 0.45 ULP, highest
error within the test set fell to 1.31 ULP, which is not perfect, but
still a significant improvement. Testing other previously erroneous
ranges no longer show particularly large accuracy errors.
Most code in newlib already uses unified syntax, but just a couple of
laggards remain. This patch removes these and means the the entire
code base has now been converted.
Having symlinks for these files led to an issue reported to the RTEMS
Project that showed up using some tar for native Windows to unpack the
newlib sources. It creates symlinks in the tar file as copies of the
files the symlinks point to. If the links appear in the tar file before
the source exists, it cannot copy the file.
The solution in this patch is to convert the files that are symbolic
links into simple files which include the file they were linked to.
This should be more portable and avoids the symbolinc link problem.
I think I may have encountered a bug in the implementation of pow:
pow(-1.0, NaN) returns 1.0 when it should return NaN.
Because ix is used to check input vs 1.0 rather than hx, -1.0 is
mistaken for 1.0
This edits licenses held by Berkeley and NetBSD, both of which
have removed the advertising requirement from their licenses.
Signed-off-by: Keith Packard <keithp@keithp.com>
This reverts commit 59362c80e3.
This breaks gnulib's autoconf test for POSIX compatibility of
fflush/fseek. After fflush/fseek, ftello and lseek are out of
sync, with lseek having the wrong offset. This breaks backward
compatibility with Cygwin applications.
Signed-off-by: Corinna Vinschen <corinna@vinschen.de>
- also change the handling of default_newlib_reent_check_verify to
be the same as other default variables in configure.host
- regenerate newlib/configure
If we had architecture-specific exception bits, we could just set them
to match the processor, but instead ieeefp.h is shared by all targets
so we need to map between the public values and the register contents.
Signed-off-by: Keith Packard <keithp@keithp.com>
This makes the fpsetround function actually do something rather than
just return -1 due to the default 'fall-through' behavior of the switch
statement.
Signed-off-by: Keith Packard <keithp@keithp.com>
In the two helper functions that _dcvt calls for 'f' and 'e' mode, if
there are no digits to display after the decimal point, don't add one.
Signed-off-by: Keith Packard <keithp@keithp.com>
Leading zeros after the decimal point should not count
towards the 'ndigits' limit.
This makes gcvt match glibc and the posix gcvt man page.
Signed-off-by: Keith Packard <keithp@keithp.com>
Even if the number is really small and this means showing *no* digits.
This makes newlib match glibc, and the fcvt posix man page.
Signed-off-by: Keith Packard <keithp@keithp.com>
Jeff Johnston
Takashi Yano
Corinna Vinschen
Jon Turney
Thomas Wolff
Mike Frysinger
Christian Biesinger via Newlib
Keith Packard
Sebastian Huber
Joel Sherrill
Roger Sayle
Matt Joyce
Christoph Muellner
Maxim Blinov
Alex White
Kito Cheng
Jonathan Wakely
Dimitar Dimitrov
Richard Earnshaw
Ola Olsson
Ties Stuij
David Macek
Marcus Comstedt
Hans-Peter Nilsson
Eshan dhawan
Paul Zimmermann
Fabian Schriever
Corinna Vinschen
Ivan Grokhotov
dougm
trasz
jah
mjg
rlibby
dab
jhb
Torbjörn SVENSSON
Christophe Lyon
Jojo R
Jozef Lawrynowicz
Craig Blackmore
Aschref Ben Thabet
Szabolcs Nagy
Joseph S. Myers
Thomas Wolff
Nicolas Brunie
Anthony Green