Corinna Vinschen [Sun, 23 Dec 2018 20:36:42 +0000 (21:36 +0100)]
Cygwin: utilize FILE_DISPOSITION_POSIX_SEMANTICS
- short-circuit most code in unlink_nt since it's not necessary
anymore if FILE_DISPOSITION_POSIX_SEMANTICS is supported.
- Immediately remove O_TMPFILE from filesystem after creation.
Disable code for now because we have to implement /proc/self/fd
opening by handle first, lest linkat fails.
Various new file info class members adding important POSIX semantics
have been added with W10 1709. We may want to utilize them, so add
a matching wincaps.
Rearrange checking the W10 build number to prefer the latest builds
over the older builds. Rename wincap_10 to wincap_10_1507 for
enhanced clarity.
Corinna Vinschen [Sat, 22 Dec 2018 23:17:34 +0000 (00:17 +0100)]
Cygwin: ntdll.h: Update _FILE_INFORMATION_CLASS
- Add missing members added in later OS versions
- Rearrange accompanying FILE_foo_INFORMATION structs
ordered by info class
- Add promising FILE_foo_INFORMATION structs of later
Windows 10 releases plus accompanying enums
- Drop "Checked on 64 bit" comments since that's self-evident
these days
Corinna Vinschen [Wed, 19 Dec 2018 20:10:37 +0000 (21:10 +0100)]
Cygwin: fix heap allocation on WOW64 and /3GB enabled 32 bit machines
The check for the TEB being allocated beyond the first 2GB area is not
valid anymore. At least on W10 WOW64, the TEB is allocated in the
lower 2GB even in large-address aware executables. Use VirtualQuery
instead. It fails for invalid addresses so that's a simple enough test.
Szabolcs Nagy [Mon, 10 Dec 2018 14:40:01 +0000 (14:40 +0000)]
Fix powf overflow handling in non-nearest rounding mode
The threshold value at which powf overflows depends on the rounding mode
and the current check did not take this into account. So when the result
was rounded away from zero it could become infinity without setting
errno to ERANGE.
Example: pow(0x1.7ac7cp+5, 23) is 0x1.fffffep+127 + 0.1633ulp
If the result goes above 0x1.fffffep+127 + 0.5ulp then errno is set,
which is fine in nearest rounding mode, but
powf(0x1.7ac7cp+5, 23) is inf in upward rounding mode
powf(-0x1.7ac7cp+5, 23) is -inf in downward rounding mode
and the previous implementation did not set errno in these cases.
The fix tries to avoid affecting the common code path or calling a
function that may introduce a stack frame, so float arithmetics is used
to check the rounding mode and the threshold is selected accordingly.
imp [Tue, 20 Nov 2018 07:11:23 +0000 (07:11 +0000)]
Ensure that all values of ns, us and ms work
for {n,u,m}stosbt
Integer overflows and wrong constants limited the accuracy of these
functions and created situatiosn where sbttoXs(Xstosbt(Y)) != Y. This
was especailly true in the ns case where we had millions of values
that were wrong.
Instead, used fixed constants because there's no way to say ceil(X)
for integer math. Document what these crazy constants are.
Also, use a shift one fewer left to avoid integer overflow causing
incorrect results, and adjust the equasion accordingly. Document this.
Allow times >= 1s to be well defined for these conversion functions
(at least the Xstosbt). There's too many users in the tree that they
work for >= 1s.
This fixes a failure on boot to program firmware on the mlx4
NIC. There was a msleep(1000) in the code. Prior to my recent rounding
changes, msleep(1000) worked, but msleep(1001) did not because the old
code rounded to just below 2^64 and the new code rounds to just above
it (overflowing, causing the msleep(1000) to really sleep 1ms).
A test program to test all cases will be committed shortly. The test
exaustively tries every value (thanks to bde for the test).
imp [Thu, 15 Nov 2018 16:02:13 +0000 (16:02 +0000)]
When converting ns,us,ms to sbt, return the ceil()
of the result rather than the floor(). Returning the floor means that
sbttoX(Xtosbt(y)) != y for almost all values of y. In practice, this
results in a difference of at most 1 in the lsb of the sbintime_t. This
difference is meaningless for all current users of these functions, but
is important for the newly introduced sysctl conversion routines which
implicitly rely on the transformation being idempotent.
pfg [Mon, 20 Nov 2017 19:43:44 +0000 (19:43 +0000)]
sys: further adoption of SPDX licensing ID tags.
Mainly focus on files that use BSD 3-Clause license.
The Software Package Data Exchange (SPDX) group provides a specification
to make it easier for automated tools to detect and summarize well known
opensource licenses. We are gradually adopting the specification, noting
that the tags are considered only advisory and do not, in any way,
superceed or replace the license texts.
Special thanks to Wind River for providing access to "The Duke of
Highlander" tool: an older (2014) run over FreeBSD tree was useful as a
starting point.
ian [Sat, 29 Jul 2017 17:00:23 +0000 (17:00 +0000)]
Add inline functions to convert between sbintime_t
and decimal time units. Use them in some existing code that is
vulnerable to roundoff errors.
The existing constant SBT_1NS is a honeypot, luring unsuspecting folks into
writing code such as long_timeout_ns*SBT_1NS to generate the argument for a
sleep call. The actual value of 1ns in sbt units is ~4.3, leading to a
large roundoff error giving a shorter sleep than expected when multiplying
by the trucated value of 4 in SBT_1NS. (The evil honeypot aspect becomes
clear after you waste a whole day figuring out why your sleeps return early.)
imp [Tue, 28 Feb 2017 23:42:47 +0000 (23:42 +0000)]
Renumber copyright clause 4
Renumber cluase 4 to 3, per what everybody else did when BSD granted
them permission to remove clause 3. My insistance on keeping the same
numbering for legal reasons is too pedantic, so give up on that point.
Submitted by: Jan Schaumann <jschauma@stevens.edu>
Pull Request: https://github.com/freebsd/freebsd/pull/96
While reformatting the script, backticks `` were replaced with
brackets $(). This in turn invalidated the \\( ... \\) expressions in the
sed script because backslash resolution in $() works differently from
backslash resolution in ``. Only a single backslash is valid now.
While at it, fix up the uname(2) date representation when building a
snapshot.
Corinna Vinschen [Fri, 30 Nov 2018 21:39:57 +0000 (22:39 +0100)]
Cygwin: clocks: fix a hang on pre-Windows 10 machines
when calling clocks too early in DLL init, the vtables are not correctly
set up for some reason. Calls to init() from now() fail because the init
pointer in the vtable is NULL.
Real life example is mintty which runs into a minor problem at startup,
triggering a system_printf call. Strace is another problem, it's called
the first time prior to any class initialization.
Workaround is to make sure that no virtual methods are called in an
early stage. Make init() non-virtual and convert resolution() to a
virtual method instead. Add a special non-virtual
clk_monotonic_t::strace_usecs.
While at it:
- Inline internal-only methods.
- Drop the `inited' member. Convert period/ticks_per_sec toa union.
Initialize period/ticks_per_sec via InterlockeExchange64.
- Fix GetTickCount64 usage. No, it's not returning ticks but
milliseconds since boot (unbiased).
Corinna Vinschen [Tue, 27 Nov 2018 12:47:02 +0000 (13:47 +0100)]
Cygwin: implement extensible clock interface
- Drop hires_[nm]s clocks, rename hires.h to clock.h.
- Implement clk_t class as an extensible clock class in new file clock.cc.
- Introduce get_clock(clock_id) returning a pointer to the clk_t instance
for clock_id. Provide the following methods along the lines of the former
hires classes:
Corinna Vinschen [Tue, 27 Nov 2018 12:59:37 +0000 (13:59 +0100)]
Cygwin: timers: clock_setres: make no-op
clock_setres is a questionable function only existing on QNX.
Disable the function, just return success for CLOCK_REALTIME
to maintain backward compatibility.
Corinna Vinschen [Mon, 26 Nov 2018 16:47:53 +0000 (17:47 +0100)]
Cygwin: timers: use spinlock to prime hires_ns thread-safe
The current method to make hires_ns priming thread-safe isn't
thread-safe. Rather than hoping that running the thread in
TIME_CRITICAL priority is doing the right thing, use a spinlock.
A previous commit introduced the ability to use the semi-hosting
SYS_EXIT_EXTENDED operation to libgloss, this commit adds the same
ability to the sys/arm/ backend so that building newlib only will
provide the same capabilities.
Jim Wilson [Mon, 19 Nov 2018 23:27:05 +0000 (15:27 -0800)]
Add libgloss install-pdf and install-html targets.
The toplevel makefile used by binutils/gcc/newlib/etc has install-pdf and
install-html targets, but they fail because libgloss doesn't support them.
Tested with an arm-eabi combined tree build and install, and verifying that
the install-pdf and install-html targets now work, and that the pdf and html
doc files are now in the install tree.
libgloss/
* Makefile.in (install-html, install-pdf): New.
* doc/Makefile.in (htmldir, pdfdir): New.
(porting.ps): Delete white space on blank line.
(install-pdf, install-html): New.
Enable return code with semi-hosting SYS_EXIT_EXTENDED
The _exit function currently passes -1 as a "sig" to the _kill function as an
invalid signal number so that _kill can distinguish between an abort and a
standard exit.
For boards using the SYS_EXIT_EXTENDED semi-hosting operation to return a
status code, this means that the "status" paramter to _exit is ignored and the
return code is always -1.
https://developer.arm.com/docs/100863/latest/semihosting-operations/sys_exit_extended-0x20
This patch puts shared code between _kill and _exit into a new function
_kill_shared that takes the semi-hosting "reason" to use (if semi-hosting is
available) as an argument.
For semi-hosting _kill_shared provides that "reason".
Without the "sig" argument being used to distinguish between a normal and
abnormal exit, the _exit function can provide the return code to be used if the
SYS_EXIT_EXTENDED operation is available.
Wilco Dijkstra [Thu, 8 Nov 2018 15:50:42 +0000 (15:50 +0000)]
Fix issue with dst bias in memset
This patch fixes an issue in the previous memset loop change. If the
zva size is >= 256 and there are more than 64 bytes left in the
tail, we could enter the loop and thus need to rebias dst by 32 as
well.
Since no known CPUs use this size this can't be tested natively, so I've
tested it on a simulator initialized with a large zva size.
Sebastian Huber [Wed, 7 Nov 2018 13:59:19 +0000 (14:59 +0100)]
Fix v850 target for RTEMS
Do not define __ATTRIBUTE_IMPURE_PTR__ for RTMES on the v850 target.
The previous definition lead to the following linker error in
combination with -fdata-sections:
relocation truncated to fit: R_V810_GPWLO_1 against symbol
`_global_impure_ptr' defined in .rodata._global_impure_ptr section in
libc.a(lib_a-impure.o)
relocation truncated to fit: R_V810_GPWLO_1 against symbol `_impure_ptr'
defined in .data._impure_ptr section in libc.a(lib_a-impure.o)
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Sebastian Huber [Thu, 8 Nov 2018 08:33:13 +0000 (09:33 +0100)]
RTEMS: Avoid <machine/param.h> in <sys/_cpuset.h>
The <machine/param.h> header file exposes some unrelated stuff not
covered by C or POSIX. Avoid its use in <sys/_cpuset.h> since it is
included in <rtems.h>.
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Wilco Dijkstra [Tue, 6 Nov 2018 14:42:10 +0000 (14:42 +0000)]
Adjust writeback in non-zero memset
This fixes an ineffiency in the non-zero memset. Delaying the writeback
until the end of the loop is slightly faster on some cores - this shows
~5% performance gain on Cortex-A53 when doing large non-zero memsets.
Cygwin: honor SOCK_NONBLOCK/SOCK_CLOEXEC in socket(2)/socketpair(2)
fhandler_socket_wsock::set_socket_handle calls set_flags after
setting the O_NONBLOCK/O_CLOEXEC flags, thus overwriting them.
It also turns out that fhandler_socket_wsock::init_events is called
too late. The inheritence flags are changed before creating the
socket event handling objects. Thus, inheritence flags for
those objects are wrong with SOCK_CLOEXEC.
Fix this by reordering the calls and setting the file flags through
fhandler_base::set_flags.
Stafford Horne [Fri, 2 Nov 2018 22:00:02 +0000 (07:00 +0900)]
newlib/configure.host: Set have_init_fini to no for OpenRISC
The new GCC port for OpenRISC will use the init_fini_array only and not
provide the init() and fini() functions. Disable the function usage by
default as its no longer needed.
Corinna Vinschen [Mon, 29 Oct 2018 15:12:54 +0000 (16:12 +0100)]
Cygwin: fix memory corruption/SEGV if certain socket functions fail
Regression introduced with 2.11.0:
The failure paths in socket, socketpair and accept4 functions and
methods accidentally release *unused* cygheap_fdmanip objects. The
subsequently called dtable::release method was designed to be called for
*used* cygheap_fdmanip objects only. Using them on unused objects leads
to NULL pointer member dereferencing.
Worse, the inet/local accept4 methods only release the cygheap_fdmanip
object but neglect to delete the just created fhandler_socket_* object.
Fix this by removing the erroneous release calls in the aforementioned
failure paths and delete the fhandler_socket_* object in accept4 instead.
David McFarland [Sun, 28 Oct 2018 19:22:44 +0000 (16:22 -0300)]
Cygwin: Fix cygheap corruption caused by cloned atomic buffer
The fhandler_base_overlapped::copyto clears atomic_write_buf on the
clone, but none of the derived classes were doing this. This allowed
the destructor to double-free the buffer and corrupt cygheap.
Clear atomic_write_buf in copyto of all derived classes.
Wilco Dijkstra [Thu, 18 Oct 2018 17:29:36 +0000 (17:29 +0000)]
Improve performance of strstr
v3: Add support for read ahead using strnlen, giving an additional 25% speedup
on large inputs (both short and long needles).
This patch significantly improves performance of strstr by using Sunday's
Quick-Search algorithm. Due to its simplicity it has the best average
performance of string matching algorithms on almost all inputs. It uses a
bad-character shift table to skip past mismatches.
The needle length is limited to 254 - this reduces the shift table memory
4 to 8 times, lowering preprocessing overhead and minimizing cache effects.
The limit also implies its worst-case performance is linear.
Larger needles are processed by the Two-Way algorithm. The macro AVAILABLE
has been improved to use strnlen to read the input in chunks. This results
in a 2.5 times speedup for large needles, reducing the performance drop when
the Quick-Search algorithm can't be used.
The code for 1-4 byte needles has been simplified and now uses unsigned
char. Since the optimized code relies on 8-bit chars, we defer to the
size-optimized implementation if CHAR_BIT > 8.
The performance gain of finding a set of randomly chosen words of size 8 in
256 bytes of English text is 14 times on AArch64. For longer haystacks the
gain is well over 20 times.
The size-optimized strstr has also been rewritten from scratch to improve
performance. On the same test the performance gain is 69%.
Tested against GLIBC testsuite, randomized tests and the GNULIB strstr test
(https://git.savannah.gnu.org/cgit/gnulib.git/tree/tests/test-strstr.c).
Sebastian Huber [Mon, 8 Oct 2018 11:33:08 +0000 (13:33 +0200)]
Add generic implementation of dirfd()
Use existing HAVE_OPENDIR define to determine if a generic
implementation should be provided. Cygwin for example has its own
implementation of opendir() and dirfd().
Signed-off-by: Sebastian Huber <sebastian.huber@embedded-brains.de>
Corinna Vinschen [Wed, 10 Oct 2018 16:01:22 +0000 (18:01 +0200)]
newlib: Build internal strtold code only if HAVE_LONG_DOUBLE defined
Commit fbace81684f8cbb80a2048c01dc545af247f5cb7
("Import correctly working strtold from David M. Gay.")
introduced two new files, strtorx.c and strtodg.c. The functions
are only called from strtold.c. However, while strtold.c is only
built if HAVE_LONG_DOUBLE is defined, the patch erroneously added
the two new files to GENERAL_SOURCES unconditionally.
Fix this by building both files only if HAVE_LONG_DOUBLE has been
defined.
Corinna Vinschen [Wed, 10 Oct 2018 15:53:55 +0000 (17:53 +0200)]
newlib: strtold: use __builtin_nanl to avoid libm dependency
Commit 6c212a8b7873703c4f98c6b68579b234918be83a
("Fix strtod ("nan") and strtold ("nan") returns wrong negative NaN")
introduced an unconditional dependency to nanl and, in turn, to libm.
Rather than including nanl in libc as well, just call __builtin_nanl
from here. Requires GCC 3.3 or later.
Christophe Lyon [Mon, 1 Oct 2018 15:52:42 +0000 (15:52 +0000)]
[ARM] Make _kill() a noreturn function.
AngelSWI_Reason_ReportException does not return accoring to the ARM
documentation, so it is valid to mark _kill() as noreturn. This way,
the compiler does not warn about _exit() returning a value despite
being noreturn.
While working on the strstr patch I noticed several copyright headers
of the new math functions are missing closing quotes after ``AS IS.
I've added these. Also update spellings of Arm Ltd in a few places
(but still use ARM LTD in upper case portion). Finally add SPDX
identifiers to make everything consistent.
Jeff Johnston [Tue, 25 Sep 2018 19:10:46 +0000 (15:10 -0400)]
Update Arm entry in COPYING.NEWLIB.
- per Wilco Dijkstra's patch
- update Arm copyright notice to not exclude AArch64 and the various generic
contributions made (like the new math functions). Also update the date and
change spelling to Arm Ltd.
Committed, CRIS port: fix fallout from time_t defaulting to 64 bits, part 2
It's been a while... I see the CRIS port broke with the
time_t-default-to-64-bit change, observable by a few test-cases in the gcc
fortran(!) tests failing, regressing when trying a recent newlib.
This is a two-part belt-and-suspenders change: adjust the CRIS port
gettimeofday syscall (the only one in newlib/CRIS passing a time_t or
struct timeval) to handle a userspace 64-bit time_t and secondly default
time_t to 32-bit long anyway. I considered making the local
"kernel_timeval" copy in _gettimeofday conditional on (userspace) time_t
being 64 bits, but thought it not worth bothering with the few move insns.
The effect of a 64-bit time_t is however observable as longer simulation
time when running the gcc testsuite and as bigger binaries without any
actual upside from the larger time_t size, so I thought better make the
default for this port go back to being a "long" again.
Tested by running the gcc testsuite over the three combinations of two
parts of the patch and observing the expected changes. Committed.
newlib:
* configure.host (cris, crisv32): Default to "long" time_t.
Committed, CRIS port: fix fallout from time_t defaulting to 64 bits, part 1
It's been a while... I see the CRIS port broke with the
time_t-default-to-64-bit change, observable by a few test-cases in the
gcc fortran(!) tests failing, regressing when trying a recent newlib.
This is a two-part belt-and-suspenders change: adjust the CRIS port
gettimeofday syscall (the only one in newlib/CRIS passing a time_t or
struct timeval) to handle a userspace 64-bit time_t and secondly default
time_t to 32-bit long anyway. I considered making the local
"kernel_timeval" copy in _gettimeofday conditional on (userspace) time_t
being 64 bits, but thought it not worth bothering with the few move insns.
The effect of a 64-bit time_t is however observable as longer simulation
time when running the gcc testsuite and as bigger binaries without any
actual upside from the larger time_t size, so I thought better make the
default for this port go back to being a "long" again.
Tested by running the gcc testsuite over the three combinations of two
parts of the patch and observing the expected changes. Committed.
libgloss:
Adjust for syscall and userspace having different time_t or timeval.
* cris/linunistd.h (kernel_time_t, kernel_suseconds_t, kernel_timeval):
New types.
(gettimeofday): Change the type of the first argument to be a
pointer to a struct kernel_timeval.
* cris/gensyscalls (_gettimeofday): Use an intermediate struct
kernel_timeval for the syscall and initialize the result from
that.
Cygwin: console: make sure EnumFontFamiliesExW loop isn't infinite
The current loop condition is borderline. Make sure it ends and
choose a replacement char in the unlikely case the current console
font isn't recognized at all.
Jon Beniston [Thu, 6 Sep 2018 12:53:15 +0000 (13:53 +0100)]
search: Fix Berkeley DB hash code for 16-bit targets.
hash.h: Use 32-bit type for data stored on disk, so code
works for 16 and 64-bit targets. Reduce maximum bucket size on 16-bit
targets, so it fits in available memory.
hash.c: Check bucket size isn't too big for target.
hash_buf.c: Fix overflow warning on 16-bit targets.
Keith Packard [Thu, 6 Sep 2018 04:39:40 +0000 (21:39 -0700)]
Use !__HAVE_LOCALE_INFO__ define to use _ctype_ directly [v2]
When __HAVE_LOCALE_INFO__ is not selected, directly access the
existing _ctype_ variable from __locale_ctype_ptr() and
__locale_ctype_ptr_l(), eliminating the need for any locale or reent
structure
Signed-off-by: Keith Packard <keithp@keithp.com>
v2:
locale: fix conflict with __locale_ctype_ptr macro
If we are building without __HAVE_LOCALE_INFO__, there is a
macro providing __locale_ctype_ptr which in turn fouls up this
declaration.
Keith Packard [Thu, 6 Sep 2018 04:25:01 +0000 (21:25 -0700)]
stdlib: Use __get_numeric_locale instead of __localeconv_l for decimal_point
The string/float conversion functions need to get the locale decimal
point. Instead of calling __localeconv_l (which copies locale data
into lconv form from __get_numeric_locale), use __get_numeric_locale
directly.
Keith Packard [Thu, 6 Sep 2018 04:23:46 +0000 (21:23 -0700)]
Include sys/syslimits.h in limits.h
This makes sure any system-defined limits are specified before the
defaults are checked. Without this, ARG_MAX and PATH_MAX end up
getting the default definitions from limits.h rather than the defines
from syslimits.h. This could potentially cause problems when
different files used different values for the same name.
Improve strstr performance for the common case of short needles. For a single
character strchr is best, for 2-4 characters a small loop is fastest. For these
the speedup over the Two-Way algorithm is ~10 times on large strings.
Newlib builds, the new code passes GLIBC testsuite. OK for commit?
Andy Koppe [Fri, 31 Aug 2018 11:26:02 +0000 (12:26 +0100)]
Avoid ARM SWI Seek when querying file position
Issuing an ARM semi-hosting Seek command when just querying file
position with SEEK_CUR and offset zero is unnecessary, because unlike
the lseek() Unix system call the Seek command does not actually return
the file position. For that reason, syscalls.c for ARM keeps track of
file position in the 'poslog', so we can just return that.
Moreover, since the Seek command only accepts an absolute file position,
SEEK_CUR operations are implemented by adding the relative offset to the
position in the poslog. If the host implements non-binary files with
implicit carriage return characters but doesn't discount those implicit
CRs when implementing Seek (by just mapping straight to Windows file
operations), this actually ended up wrongly changing file position when
using SEEK_CUR with offset zero or functions like ftell() or fgetpos()
that are based on that.
Commit 35998fc2fa6cbb7d761f6d88346246bd3627552b fixed the buffer underun
in win32 path normalization, but introduced a new bug: A wrong
assumption led to the inability to backtrack the path outside of the
current working directory in case of relative paths.
This patch fixes this problem, together with a minor problem if the CWD
is on a network share: The result erroneously started with tripple
backslash if the src path starts with a single backslash.
Andy Koppe [Wed, 29 Aug 2018 14:37:13 +0000 (15:37 +0100)]
Drop non-reserved parameter names from sys/signal.h
Standard headers shouldn't use non-reserved identifiers as parameter
names in function declarations, because programs could in theory
define macros with such names before including a header.
git://sourceware.org / newlib-cygwin.git / log