When build-many-glibcs.py re-execs itself with execv, any buffered
output on stdout may be lost (in particular, messages intended to go
to a bot's log about the re-exec taking place). This patch makes it
flush stdout before execv, similar to the flush before running a
subprocess from the bot that is done to ensure output appears in the
right order.
* scripts/build-many-glibcs.py (Context.exec_self): Flush stdout
before calling execv.
Current optimized powercp64/power7 memchr uses a strategy to check for
p versus align(p+n) (where 'p' is the input char pointer and n the
maximum size to check for the byte) without taking care for possible
overflow on the pointer addition in case of large 'n'.
It was triggered by 3038145ca2 where default rawmemchr (used to
created ppc64 rawmemchr in ifunc selection) now uses memchr (p, c, (size_t)-1)
on its implementation.
This patch fixes it by implement a satured addition where overflows
sets the maximum pointer size to UINTPTR_MAX.
Checked on powerpc64le-linux-gnu.
[BZ# 20971]
* sysdeps/powerpc/powerpc64/power7/memchr.S (__memchr): Avoid
overflow in pointer addition.
* string/test-memchr.c (do_test): Add an argument to pass as
the size on memchr.
(test_main): Add check for SIZE_MAX.
This patch converts the wrapper scalbln (which set errno directly
rather than doing anything with __kernel_standard) to use the
type-generic template machinery, in the same way that has been done
for ldexp.
Tested for powerpc64le, s390, and x86_64.
Various fmax and fmin function implementations mishandle sNaN
arguments:
(a) When both arguments are NaNs, the return value should be a qNaN,
but sometimes it is an sNaN if at least one argument is an sNaN.
(b) Under TS 18661-1 semantics, if either argument is an sNaN then the
result should be a qNaN (whereas if one argument is a qNaN and the
other is not a NaN, the result should be the non-NaN argument).
Various implementations treat sNaNs like qNaNs here.
This patch fixes the x86 and x86_64 versions (ignoring float and
double for 32-bit x86 given the inability to reliably avoid the sNaN
turning into a qNaN before it gets to the called function). Tests of
sNaN inputs to these functions are added.
Note on architecture versions I haven't changed for this issue:
AArch64 already gets this right (it uses a hardware instruction with
the correct semantics for both quiet and signaling NaNs) and does not
need changes. It's possible Alpha, IA64, SPARC might need changes
(this would be shown by the testsuite if so).
Tested for x86_64 and x86 (both i686 and i586 builds, to cover the
different x86 implementations).
[BZ #20947]
* sysdeps/i386/fpu/s_fmaxl.S (__fmaxl): Add the arguments when
either is a signaling NaN.
* sysdeps/i386/fpu/s_fminl.S (__fminl): Likewise. Make code
follow fmaxl more closely.
* sysdeps/i386/i686/fpu/s_fmaxl.S (__fmaxl): Add the arguments
when either is a signaling NaN.
* sysdeps/i386/i686/fpu/s_fminl.S (__fminl): Likewise.
* sysdeps/x86_64/fpu/s_fmax.S (__fmax): Likewise.
* sysdeps/x86_64/fpu/s_fmaxf.S (__fmaxf): Likewise.
* sysdeps/x86_64/fpu/s_fmaxl.S (__fmaxl): Likewise.
* sysdeps/x86_64/fpu/s_fmin.S (__fmin): Likewise.
* sysdeps/x86_64/fpu/s_fminf.S (__fminf): Likewise.
* sysdeps/x86_64/fpu/s_fminl.S (__fminl): Likewise.
* math/libm-test.inc (fmax_test_data): Add tests of sNaN inputs.
(fmin_test_data): Likewise.
Various fmax and fmin function implementations mishandle sNaN
arguments:
(a) When both arguments are NaNs, the return value should be a qNaN,
but sometimes it is an sNaN if at least one argument is an sNaN.
(b) Under TS 18661-1 semantics, if either argument is an sNaN then the
result should be a qNaN (whereas if one argument is a qNaN and the
other is not a NaN, the result should be the non-NaN argument).
Various implementations treat sNaNs like qNaNs here.
This patch fixes the powerpc versions of these functions (shared by
float and double, 32-bit and 64-bit). The structure of those versions
is that all ordered cases are already handled before anything dealing
with the case where the arguments are unordered; thus, this patch
causes no change to the code executed in the common case (neither
argument a NaN).
Tested for powerpc (32-bit and 64-bit), together with tests to be
added along with the x86_64 / x86 fixes.
[BZ #20947]
* sysdeps/powerpc/fpu/s_fmax.S (__fmax): Add the arguments when
either is a signaling NaN.
* sysdeps/powerpc/fpu/s_fmin.S (__fmin): Likewise.
Various fmax and fmin function implementations mishandle sNaN
arguments:
(a) When both arguments are NaNs, the return value should be a qNaN,
but sometimes it is an sNaN if at least one argument is an sNaN.
(b) Under TS 18661-1 semantics, if either argument is an sNaN then the
result should be a qNaN (whereas if one argument is a qNaN and the
other is not a NaN, the result should be the non-NaN argument).
Various implementations treat sNaNs like qNaNs here.
This patch fixes the generic implementations used in the absence of
architecture-specific versions.
Tested for mips64 and powerpc (together with testcases that I'll add
along with the x86_64 / x86 fixes).
[BZ #20947]
* math/s_fmax_template.c (M_DECL_FUNC (__fmax)): Add the arguments
when either is a signaling NaN.
* math/s_fmin_template.c (M_DECL_FUNC (__fmin)): Likewise.
Information about whether the ABI of long double is the same as that
of double is split between bits/mathdef.h and bits/wordsize.h.
When the ABIs are the same, bits/mathdef.h defines
__NO_LONG_DOUBLE_MATH. In addition, in the case where the same glibc
binary supports both -mlong-double-64 and -mlong-double-128,
bits/wordsize.h defines __LONG_DOUBLE_MATH_OPTIONAL, along with
__NO_LONG_DOUBLE_MATH if this particular compilation is with
-mlong-double-64.
As part of the refactoring I proposed in
<https://sourceware.org/ml/libc-alpha/2016-11/msg00745.html>, this
patch puts all that information in a single header,
bits/long-double.h. It is included from sys/cdefs.h alongside the
include of bits/wordsize.h, so other headers generally do not need to
include bits/long-double.h directly.
Previously, various bits/mathdef.h headers and bits/wordsize.h headers
had this long double information (including implicitly in some
bits/mathdef.h headers through not having the defines present in the
default version). After the patch, it's all in six bits/long-double.h
headers. Furthermore, most of those new headers are not
architecture-specific. Architectures with optional long double all
use the ldbl-opt sysdeps directory, either in the order (ldbl-64-128,
ldbl-opt, ldbl-128) or (ldbl-128ibm, ldbl-opt). Thus a generic header
for the case where long double = double, and headers in ldbl-128,
ldbl-96 and ldbl-opt, suffices to cover every architecture except for
cases where long double properties vary between different ABIs sharing
a set of installed headers; fortunately all the ldbl-opt cases share a
single compiler-predefined macro __LONG_DOUBLE_128__ that can be used
to tell whether this compilation is -mlong-double-64 or
-mlong-double-128.
The two cases where a set of headers is shared between ABIs with
different long double properties, MIPS (o32 has long double = double,
other ABIs use ldbl-128) and SPARC (32-bit has optional long double,
64-bit has required long double), need their own bits/long-double.h
headers.
As with bits/wordsize.h, multiple-include protection for this header
is generally implicit through the include guards on sys/cdefs.h, and
multiple inclusion is harmless in any case. There is one subtlety:
the header must not define __LONG_DOUBLE_MATH_OPTIONAL if
__NO_LONG_DOUBLE_MATH was defined before its inclusion, because doing
so breaks how sysdeps/ieee754/ldbl-opt/nldbl-compat.h defines
__NO_LONG_DOUBLE_MATH itself before including system headers. Subject
to keeping that working, it would be reasonable to move these macros
from defined/undefined #ifdef to always-defined 1/0 #if semantics, but
this patch does not attempt to do so, just rearranges where the macros
are defined.
After this patch, the only use of bits/mathdef.h is the alpha one for
modifying complex function ABIs for old GCC. Thus, all versions of
the header other than the default and alpha versions are removed, as
is the include from math.h.
Tested for x86_64 and x86. Also did compilation-only testing with
build-many-glibcs.py.
* bits/long-double.h: New file.
* sysdeps/ieee754/ldbl-128/bits/long-double.h: Likewise.
* sysdeps/ieee754/ldbl-96/bits/long-double.h: Likewise.
* sysdeps/ieee754/ldbl-opt/bits/long-double.h: Likewise.
* sysdeps/mips/bits/long-double.h: Likewise.
* sysdeps/unix/sysv/linux/sparc/bits/long-double.h: Likewise.
* math/Makefile (headers): Add bits/long-double.h.
* misc/sys/cdefs.h: Include <bits/long-double.h>.
* stdlib/strtold.c: Include <bits/long-double.h> instead of
<bits/wordsize.h>.
* bits/mathdef.h [!_COMPLEX_H]: Do not allow inclusion.
[!__NO_LONG_DOUBLE_MATH]: Remove conditional code.
* math/math.h: Do not include <bits/mathdef.h>.
* sysdeps/aarch64/bits/mathdef.h: Remove file.
* sysdeps/alpha/bits/mathdef.h [!_COMPLEX_H]: Do not allow
inclusion.
* sysdeps/ia64/bits/mathdef.h: Remove file.
* sysdeps/m68k/m680x0/bits/mathdef.h: Likewise.
* sysdeps/mips/bits/mathdef.h: Likewise.
* sysdeps/powerpc/bits/mathdef.h: Likewise.
* sysdeps/s390/bits/mathdef.h: Likewise.
* sysdeps/sparc/bits/mathdef.h: Likewise.
* sysdeps/x86/bits/mathdef.h: Likewise.
* sysdeps/s390/s390-32/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]: Remove
conditional code.
* sysdeps/s390/s390-64/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]:
Likewise.
* sysdeps/unix/sysv/linux/alpha/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]:
Likewise.
* sysdeps/unix/sysv/linux/powerpc/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]:
Likewise.
* sysdeps/unix/sysv/linux/sparc/bits/wordsize.h
[!__NO_LONG_DOUBLE_MATH && !__LONG_DOUBLE_MATH_OPTIONAL]:
Likewise.
This patch makes bits/fcntl-linux.h include <linux/falloc.h> to define
the FALLOC_* flags under __USE_GNU (linux/falloc.h defines only those
bits, nothing else).
Tested for x86_64 and x86.
* sysdeps/unix/sysv/linux/bits/fcntl-linux.h [__USE_GNU]: Include
<linux/falloc.h>.
(FALLOC_FL_KEEP_SIZE): Remove.
(FALLOC_FL_PUNCH_HOLE): Likewise.
(FALLOC_FL_COLLAPSE_RANGE): Likewise.
(FALLOC_FL_ZERO_RANGE): Likewise.
The macro ENTRY in tst-strfrom.h is used to generate the input values for
each floating-point type (float, double, long double). It should append
the parameter LSUF (Literal suffix) to the floating-point number, but is
using CSUF (C function suffix). This patch fixes it.
Tested for powerpc64le and x86_64.
This patch consolidates the Linux renameat implementation on
sysdeps/unix/sysv/linux/renameat.c. The renameat syscall was
deprecated at b0da6d44 for newer architectures, so using the
auto-generation list may generate wrappers that returns ENOSYS.
Current code try to use __NR_renameat and if it is not define
it uses __NR_renameat2.
Checked on x86_64 and aarch64.
* sysdeps/unix/sysv/linux/renameat.c: New file.
* sysdeps/unix/sysv/linux/syscalls.list: Remove renameat.
This patch consolidates the Linux rename implementation on
sysdeps/unix/sysv/linux/rename.c. Current code try to use
__NR_rename if is defined and apply the same strategy for
__NR_renameat and __NR_renameat2.
Check on x86_64 and aarch64.
* sysdeps/unix/sysv/linux/rename.c: New file.
* sysdeps/unix/sysv/linux/generic/rename.c: Remove file.
calls strcspn, call strcspn directly so we get the end of the token without
an extra call to rawmemchr. Also avoid an unnecessary call to strcspn after
the last token by adding an early exit for an empty string. Change strtok
to tailcall strtok_r to avoid unnecessary code duplication.
Remove the special header optimization for strtok_r of a 1-character
constant string - both strspn and strcspn contain optimizations for this
case. Benchmarking this showed similar performance in the worst case,
but up to 5.5x better performance in the "found" case for large inputs.
* benchtests/bench-strtok.c (oldstrtok): Add old implementation.
* string/strtok.c (strtok): Change to tailcall __strtok_r.
* string/strtok_r.c (__strtok_r): Optimize for performance.
* string/string-inlines.c (__old_strtok_r_1c): New function.
* string/bits/string2.h (__strtok_r): Move to string-inlines.c.
This patch converts the wrapper log1p (which set errno directly rather
than doing anything with __kernel_standard) to use the type-generic
template machinery, in the same way that has been done for ilogb.
Tested for powerpc64le, s390, and x86_64.
assembler version by tailcalling memchr with the maximum size.
If a target has an optimized memchr this is significantly faster,
if not, then this makes little difference.
Also optimize the special case of zero to use strlen as this is
typically faster than memchr.
* string/rawmemchr.c (RAWMEMCHR): Use faster memchr/strlen.
This patch updates build-many-glibcs.py to use Linux 4.9 for kernel
headers unless another version is explicitly specified. Note that
when a version changes like this you'll need to use --replace-sources
when updating an existing checkout to tell build-many-glibcs.py it's
OK to delete and replace the sources of a component for which the
version used has changed.
* scripts/build-many-glibcs.py (Context.checkout): Default Linux
kernel version to 4.9.
The new test framework changed the call stacks. The weak attribute
acts as a strong compiler barrier. Some static functions had to be
declared extern as a consequence.
The new test driver in <support/test-driver.c> has feature parity with
the old one. The main difference is that its hooking mechanism is
based on functions and function pointers instead of macros. This
commit also implements a new environment variable, TEST_COREDUMPS,
which disables the code which disables coredumps (that is, it enables
them if the invocation environment has not disabled them).
<test-skeleton.c> defines wrapper functions so that it is possible to
use existing macros with the new-style hook functionality.
This commit changes only a few test cases to the new test driver, to
make sure that it works as expected.
For many years, the only effect of these macros has been to make
unistd.h declare getlogin_r. _POSIX_C_SOURCE >= 199506L also causes
this function to be declared. However, people who don't carefully
read all the headers might be confused into thinking they need to
define _REENTRANT for any threaded code (as was indeed the case a long
time ago).
Therefore, remove __USE_REENTRANT, and make _REENTRANT and _THREAD_SAFE
into synonyms for _POSIX_C_SOURCE=199506L. This will only affect
programs that don't select a higher conformance level some other way.
For instance, -std=c89 -D_REENTRANT will see a change in visible
declarations, but -std=c99 -D_POSIX_C_SOURCE=200809L -D_REENTRANT won't,
and -D_REENTRANT all by itself also won't, because _DEFAULT_SOURCE
implies _POSIX_C_SOURCE > 199506.
* include/features.h: Remove __USE_REENTRANT. Treat _REENTRANT
and _THREAD_SAFE the same as _POSIX_C_SOURCE=199506L, if a higher
POSIX conformance level has not been selected by other macros.
* NEWS, manual/creature.texi: Document this change.
* posix/unistd.h, posix/bits/unistd.h: Don't check __USE_REENTRANT.
* include/libc-symbols.h: Don't define _REENTRANT.
* scripts/check-installed-headers.sh: Don't undefine _REENTRANT.
The pretty-printers changes resulted in the new tests showing as
UNRESOLVED (missing .out files) when run-built-tests = no, so
resulting in "make check" exiting with error status.
https://sourceware.org/ml/libc-testresults/2016-q4/msg00014.html
This patch moves the use of $(tests-printers) when generating the
summary of results into the setting of tests-expected, which is
appropriately conditional, matching the conditional for when tests
depends on $(tests-printers-out).
Tested with cross to aarch64 with build-many-glibcs.py, and natively
on x86_64.
* Rules [$(run-built-tests) != no] (tests-expected): Add
$(tests-printers).
(tests): Do not pass $(tests-printers) to merge-test-results.sh.
This patch adds pretty printers for the following NPTL types:
- pthread_mutex_t
- pthread_mutexattr_t
- pthread_cond_t
- pthread_condattr_t
- pthread_rwlock_t
- pthread_rwlockattr_t
To load the pretty printers into your gdb session, do the following:
python
import sys
sys.path.insert(0, '/path/to/glibc/build/nptl/pretty-printers')
end
source /path/to/glibc/source/pretty-printers/nptl-printers.py
You can check which printers are registered and enabled by issuing the
'info pretty-printer' gdb command. Printers should trigger automatically when
trying to print a variable of one of the types mentioned above.
The printers are architecture-independent, and were tested on an AMD64 running
Ubuntu 14.04 and an x86 VM running Fedora 24.
In order to work, the printers need to know the values of various flags that
are scattered throughout pthread.h and pthreadP.h as enums and #defines. Since
replicating these constants in the printers file itself would create a
maintenance burden, I wrote a script called gen-py-const.awk that Makerules uses
to extract the constants. This script is pretty much the same as gen-as-const.awk,
except it doesn't cast the constant values to 'long' and is thorougly documented.
The constants need only to be enumerated in a .pysym file, which is then referenced
by a Make variable called gen-py-const-headers.
As for the install directory, I discussed this with Mike Frysinger and Siddhesh
Poyarekar, and we agreed that it can be handled in a separate patch, and shouldn't
block merging of this one.
In addition, I've written a series of test cases for the pretty printers.
Each lock type (mutex, condvar and rwlock) has two test programs, one for itself
and other for its related 'attributes' object. Each test program in turn has a
PExpect-based Python script that drives gdb and compares its output to the
expected printer's. The tests run on the glibc host, which is assumed to have
both gdb and PExpect; if either is absent the tests will fail with code 77
(UNSUPPORTED). For cross-testing you should use cross-test-ssh.sh as test-wrapper.
I've tested the printers on both native builds and a cross build using a Beaglebone
Black running Debian, with the build system's filesystem shared with the board
through NFS.
Finally, I've written a README that explains all this and more.
* INSTALL: Regenerated.
* Makeconfig: Add comments and whitespace to make the control flow
clearer.
(+link-printers-tests, +link-pie-printers-tests, CFLAGS-printers-tests,
installed-rtld-LDFLAGS, built-rtld-LDFLAGS, link-libc-rpath,
link-libc-tests-after-rpath-link, link-libc-printers-tests): New.
(rtld-LDFLAGS, rtld-tests-LDFLAGS, link-libc-tests-rpath-link,
link-libc-tests): Use the new variables as required.
* Makerules ($(py-const)): New rule.
generated: Add $(py-const).
* README.pretty-printers: New file.
* Rules (tests-printers-programs, tests-printers-out, py-env): New.
(others): Depend on $(py-const).
(tests): Depend on $(tests-printers-programs) or $(tests-printers-out),
as required. Pass $(tests-printers) to merge-test-results.sh.
* manual/install.texi: Add requirements for testing the pretty printers.
* nptl/Makefile (gen-py-const-headers, pretty-printers, tests-printers,
CFLAGS-test-mutexattr-printers.c CFLAGS-test-mutex-printers.c,
CFLAGS-test-condattr-printers.c, CFLAGS-test-cond-printers.c,
CFLAGS-test-rwlockattr-printers.c CFLAGS-test-rwlock-printers.c,
tests-printers-libs): Define.
* nptl/nptl-printers.py: New file.
* nptl/nptl_lock_constants.pysym: Likewise.
* nptl/test-cond-printers.c: Likewise.
* nptl/test-cond-printers.py: Likewise.
* nptl/test-condattr-printers.c: Likewise.
* nptl/test-condattr-printers.py: Likewise.
* nptl/test-mutex-printers.c: Likewise.
* nptl/test-mutex-printers.py: Likewise.
* nptl/test-mutexattr-printers.c: Likewise.
* nptl/test-mutexattr-printers.py: Likewise.
* nptl/test-rwlock-printers.c: Likewise.
* nptl/test-rwlock-printers.py: Likewise.
* nptl/test-rwlockattr-printers.c: Likewise.
* nptl/test-rwlockattr-printers.py: Likewise.
* scripts/gen-py-const.awk: Likewise.
* scripts/test_printers_common.py: Likewise.
* scripts/test_printers_exceptions.py: Likewise.
This patch adds a --strip option to build-many-glibcs.py, to make it
strip the installed shared libraries after installation. This is for
convenience if you want to compare installed stripped shared libraries
before and after a patch that was not meant to result in any code
changes: you can run with this option, copy the install/glibcs
directory, run again with the patch and compare the */lib*
subdirectory contents.
(It might make sense for the option to strip libraries in other
directories, including stripping debug information from static
libraries, with a view to making it possible for a
no-generated-code-changes patch to result in completely identical
install/glibcs directories, so simplifying comparison, though that may
need other build determinism changes, e.g. to build deterministic .a
files.)
* scripts/build-many-glibcs.py (Context.__init__): Take strip
argument.
(Glibc.build_glibc): Strip installed shared libraries if
requested.
(get_parser): Add --strip option.
(main): Update Context call.
Commit 6c9e1be87a wrongly fixes BZ#20847 by lefting the else branch
on maybe_script_execute to still being able to invalid write on stack
allocated buffer. It happens if execvp{e} is executed with an empty
arguments list ({ NULL }) and although manual states first argument
should be the script name itself, by convention, old and current
implementation allows it.
This patch fixes the issue by just account for arguments and not the
final 'NULL' (since the 'argv + 1' will indeed ignored the script name).
The empty argument list is handled in a special case with a minimum
allocated size. The patch also adds extra tests for such case in
tst-vfork3.
Tested on x86_64.
[BZ #20847]
* posix/execvpe.c (maybe_script_execute): Remove write past allocated
array bounds for else branch.
(__execvpe): Style fixes.
* posix/tst-vfork3.c (run_script): New function.
(create_script): Likewise.
(do_test): Use run_script internal function.
(do_prepare): Use create_script internal function.
TS 18661-1 generally defines libm functions taking sNaN arguments to
return qNaN and raise "invalid", even for the cases where a
corresponding qNaN argument would not result in a qNaN return. This
includes hypot with one argument being an infinity and the other being
an sNaN. This patch duly fixes hypot implementatations in glibc
(generic and powerpc) to ensure qNaN, computed by arithmetic on the
arguments, is returned in that case.
Various implementations do their checks for infinities and NaNs inline
by manipulating the representations of the arguments. For simplicity,
this patch just uses issignaling to check for sNaN arguments. This
could be inlined like the existing code (with due care about reversed
quiet NaN conventions, for implementations where that is relevant),
but given that all these checks are in cases where it's already known
at least one argument is not finite, which should be the uncommon
case, that doesn't seem worthwhile unless performance issues are
observed in practice.
Tested for x86_64, x86, mips64 and powerpc.
[BZ #20940]
* sysdeps/ieee754/dbl-64/e_hypot.c (__ieee754_hypot): Do not
return Inf for arguments Inf and sNaN.
* sysdeps/ieee754/flt-32/e_hypotf.c (__ieee754_hypotf): Likewise.
* sysdeps/ieee754/ldbl-128/e_hypotl.c (__ieee754_hypotl):
Likewise.
* sysdeps/ieee754/ldbl-128ibm/e_hypotl.c (__ieee754_hypotl):
Likewise.
* sysdeps/ieee754/ldbl-96/e_hypotl.c (__ieee754_hypotl): Likewise.
* sysdeps/powerpc/fpu/e_hypot.c (TEST_INF_NAN): Do not return Inf
for arguments Inf and sNaN. When returning a NaN, compute it by
arithmetic on the arguments.
* sysdeps/powerpc/fpu/e_hypotf.c (TEST_INF_NAN): Likewise.
* math/libm-test.inc (pow_test_data): Add tests of sNaN arguments.
The x86_64/x86 powl implementations mishandle sNaN arguments, both by
returning sNaN in some cases (instead of doing arithmetic on the
arguments to produce the result when NaN arguments result in NaN
results) and by treating sNaN the same as qNaN for arguments (1, sNaN)
and (sNaN, 0), contrary to TS 18661-1 which requires those cases to
return qNaN instead of 1.
This patch makes the x86_64/x86 powl implementations follow TS 18661-1
semantics for sNaN arguments; sNaN tests are also added for pow.
Given the problems with testing float and double sNaN arguments on
32-bit x86 (sNaN tests disabled because the compiler may convert
unnecessarily to a qNaN when passing arguments), no changes are made
to the powf and pow implementations there.
Tested for x86_64 and x86.
[BZ #20916]
* sysdeps/i386/fpu/e_powl.S (__ieee754_powl): Do not return 1 for
arguments (sNaN, 0) or (1, sNaN). Do arithmetic on NaN arguments
to compute result.
* sysdeps/x86_64/fpu/e_powl.S (__ieee754_powl): Likewise.
* math/libm-test.inc (pow_test_data): Add tests of sNaN arguments.
This uses atomic operations to access lock elision metadata that is accessed
concurrently (ie, adapt_count fields). The size of the data is less than a
word but accessed only with atomic loads and stores; therefore, we add
support for shorter-size atomic load and stores too.
* include/atomic.h (__atomic_check_size_ls): New.
(atomic_load_relaxed, atomic_load_acquire, atomic_store_relaxed,
atomic_store_release): Use it.
* sysdeps/x86/elide.h (ACCESS_ONCE): Remove.
(elision_adapt, ELIDE_LOCK): Use atomics.
* sysdeps/unix/sysv/linux/x86/elision-lock.c (__lll_lock_elision): Use
atomics and improve code comments.
* sysdeps/unix/sysv/linux/x86/elision-trylock.c
(__lll_trylock_elision): Likewise.
Various pow function implementations mishandle sNaN arguments in
various ways. This includes returning sNaN instead of qNaN for sNaN
arguments. For arguments (1, sNaN) and (sNaN, 0), TS 18661-1
semantics are also that the result should be qNaN, whereas with a qNaN
argument there the result should be 1, but for the dbl-64
implementation of pow there are issues with sNaN arguments beyond not
implementing the TS 18661-1 semantics in those special cases.
This patch makes the implementations in sysdeps/ieee754 follow the TS
18661-1 semantics consistently. Because x86 / x86_64 implementations
still need fixing, testcases are not included with this patch; they
will be included with the fix for the x86 / x86_64 versions.
Tested for x86_64, x86, mips64 and powerpc (with such testcases, which
pass in the mips64 and powerpc cases).
[BZ #20916]
* sysdeps/ieee754/dbl-64/e_pow.c (__ieee754_pow): Do not return 1
for arguments (sNaN, 0) or (1, sNaN). Do arithmetic on NaN
arguments to compute result.
* sysdeps/ieee754/flt-32/e_powf.c (__ieee754_powf): Do not return
1 for arguments (sNaN, 0) or (1, sNaN).
* sysdeps/ieee754/ldbl-128/e_powl.c (__ieee754_powl): Likewise.
* sysdeps/ieee754/ldbl-128ibm/e_powl.c (__ieee754_powl): Likewise.
The dbl-64 implementation of __ieee754_pow returns a NaN for pow
(qNaN, 0) when it should return 1. Normally this is covered up by the
wrappers ending up calling __kernel_standard which fixes up the result
for this case, but for -lieee the wrappers are bypassed and the bad
result gets through as a return value.
Now, the wrappers fixing this are dealing with variant error handling
that wants a result of NaN for pow (qNaN, 0), and only ever call
__kernel_standard for this case if NaN resulted from __ieee754_pow.
This leads to a question of whether the dbl-64 code might be
deliberately returning NaN in order to use those code paths. However,
I can find no sign that this is deliberate. If it were deliberate one
would expect other implementations to do the same, and would expect
the return of NaN to be very old, but it appears it came in by
accident when the present e_pow.c implementation replaced an fdlibm
implementation in 2001. So it appears to be unintended that this path
through the pow wrapper could be used at all.
So this patch fixes the implementation to return 1 in this case as
expected. This is consistent with all the other implementations. The
relevant path through the wrappers is now unreachable, so is removed
(which is the main motivation of this patch: to avoid that path
becoming accidentally reachable when implementing TS 18661-1 semantics
that pow (sNaN, 0) should return qNaN with "invalid" raised). Another
path that would require __ieee754_pow (0, 0) to return 0 is also
unreachable (as all implementations return 1, in accordance with C99
semantics), so is removed as well.
Note: we don't have anything set up to test -lieee, which in any case
is obsolescent (at some point we should remove the ability for new
programs to access _LIB_VERSION or define matherr and have it called
by glibc). So testing will be implicit through sNaN tests added when
making sNaN inputs work correctly for pow functions.
Tested for x86_64 and x86.
[BZ #20919]
* sysdeps/ieee754/dbl-64/e_pow.c (__ieee754_pow): Do not return
NaN first argument when raised to power 0.
* math/w_pow.c (__pow): Do not check for NaN or zero results from
raising to power zero.
* math/w_powf.c (__powf): Likewise.
* math/w_powl.c (__powl): Likewise.
* sysdeps/ieee754/k_standard.c (__kernel_standard): Do not handle
pow (0, 0) or pow (NaN, 0).
Some configurations may use NSS cryptographic routines but have no
static library for those routines. The following changes allow glibc to
be built and tested with --enable-nss-crypt, but without having a static
NSS library. At a high level the change does two things:
(1) Detect at configure time if static NSS crypto libraries are
available. Assumes libfreebl3.a (instead of the existing Fedora
libfreebl.a which is incomplete) which matches libfreebl3.so.
(2) If static NSS crypto libraries are _not_ available then adjust the
way in which we build tst-linkall-static. This includes excluding a
reference to crypt and not linking against libcrypt.a, all of which
will fail otherwise.
Testing assumptions:
* Static library is named libfreebl3.a (not libfreebl.a as is currently
provided in Fedora), matching libfreebl3.so shared link name.
Tested on x86_64 on Fedora with:
(a) --enable-nss-crypt, with no static NSS library support: PASS
(previous FAIL)
(b) --enable-nss-crypt, with faked static NSS library support: PASS
(unsupported)
* Requires changing elf/Makefile to include a stub
/lib64/libfreebl3.a for testing purposes.
(c) --disable-nss-crypt: PASS
(default)
No regressions on x86_64.
For details see:
https://www.sourceware.org/ml/libc-alpha/2016-11/msg00647.html
The first dlopen ("tst-latepthreadmod.so", RTLD_LOCAL | RTLD_LAZY) call
in elf/tst-latepthread.c fails on s390x with "error: dlopen failed:
.../build-s390x/elf/tst-latepthreadmod.so:
undefined symbol: this_function_is_not_defined".
In elf/tst-latepthreadmod.c, this_function_is_not_defined is a sibling
call which leads to a R_390_GLOB_DAT relocation in .rela.dyn instead of
a R_390_JMP_SLOT in .rela.plt.
As RTLD_LAZY skips the JMP_SLOT relocations, but not GLOB_DAT ones,
the dlopen call fails. If elf/tst-latepthreadmod.c is build with
-fno-optimize-sibling-calls, a JMP_SLOT relocation is generated for
this_function_is_not_defined and the test passes.
ChangeLog:
* elf/Makefile (CFLAGS-tst-latepthreadmod.c):
Add -fno-optimize-sibling-calls.
TS 18661-1 defines llogb functions that are like ilogb except that
they return long int instead of int. Corresponding FP_LLOGB* macros
are defined, whose values are required to have the obvious
correspondence to those of the FP_ILOGB* macros.
This patch implements these functions and macros for glibc. llogb
uses the type-generic infrastructure, with an implementation similar
to the wrapper for ilogb but with additional conversion from FP_ILOGB*
to FP_LLOGB*; this approach avoids needing to modify or duplicate any
of the architecture-specific ilogb implementations. Tests are also
based on those for ilogb.
Ideally the llogb functions would alias the ilogb ones when long is
32-bit, but such aliasing requires the associated header declarations
of the different-type alias to be hidden, typically by defining macros
before including the header (see e.g. how
sysdeps/ieee754/dbl-64/wordsize-64/s_llround.c defines lround to
__hidden_lround before including <math.h>). The infrastructure for
type-generic function implementations does not support defining such
macros at present (since C code can't define a macro whose name is
determined by other macros). So this patch leaves them as separate
functions (similar to e.g. scalbln and scalbn being separate in such a
case as well), but with the remapping of FP_ILOGB* to FP_LLOGB*
conditioned out in the case where it would be the identity map.
Tested for x86_64, x86, mips64 and powerpc.
* math/bits/mathcalls.h [__GLIBC_USE (IEC_60559_BFP_EXT)] (llogb):
New declaration.
* math/tgmath.h [__GLIBC_USE (IEC_60559_BFP_EXT)] (llogb): New
macro.
* math/math.h [__GLIBC_USE (IEC_60559_BFP_EXT)] (__FP_LONG_MAX):
New macro.
[__GLIBC_USE (IEC_60559_BFP_EXT)] (FP_LLOGB0): Likewise.
[__GLIBC_USE (IEC_60559_BFP_EXT)] (FP_LLOGBNAN): Likewise.
* math/Versions (llogb): New libm symbol at version GLIBC_2.25.
(llogbf): Likewise.
(llogbl): Likewise.
* math/Makefile (gen-libm-calls): Add w_llogbF.
(tests): Add test-fp-llogb-constants.
* math/w_llogb_template.c: New file. Based on
math/w_ilogb_template.c.
* math/libm-test.inc (llogb_test_data): New array.
(llogb_test): New function.
(main): Call llogb_test.
* math/test-fp-llogb-constants.c: New file. Based on
math/test-fp-ilogb-constants.c.
* math/test-tgmath-ret.c (llogb): New CHECK_RET_CONST call.
(do_test): Call check_return_llogb.
* math/test-tgmath.c (NCALLS): Increase to 126.
(F(compile_test)): Call llogb.
(F(llogb)): New function.
* manual/math.texi (Exponents and Logarithms): Document llogb,
llogbf, llogbl, FP_LLOGB0 and FP_LLOGBNAN.
* manual/libm-err-tab.pl (@all_functions): Add llogb.
* sysdeps/ieee754/ldbl-opt/nldbl-llogb.c: New file.
* sysdeps/ieee754/ldbl-opt/w_llogbl.c: Likewise.
* sysdeps/ieee754/ldbl-opt/Makefile (libnldbl-calls): Add llogb.
(CFLAGS-nldbl-llogb.c): New variable.
* sysdeps/nacl/libm.abilist: Update.
* sysdeps/unix/sysv/linux/aarch64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/alpha/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/arm/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/hppa/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/i386/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/ia64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/m68k/coldfire/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/m68k/m680x0/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/microblaze/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/mips/mips32/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/mips/mips64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/nios2/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/fpu/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc32/nofpu/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/libm-le.abilist:
Likewise.
* sysdeps/unix/sysv/linux/powerpc/powerpc64/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/s390/s390-32/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/s390/s390-64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/sh/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/sparc/sparc32/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/sparc/sparc64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/tile/tilegx/tilegx32/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/tile/tilegx/tilegx64/libm.abilist:
Likewise.
* sysdeps/unix/sysv/linux/tile/tilepro/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/x86_64/64/libm.abilist: Likewise.
* sysdeps/unix/sysv/linux/x86_64/x32/libm.abilist: Likewise.
When I added the getpayload and setpayload functions I failed to make
the ldbl-128 functions use the _Float128 type name like most other
ldbl-128 functions do in preparation for being used to implement *f128
functions. This patch fixes them to use that name.
Tested for mips64.
* sysdeps/ieee754/ldbl-128/s_getpayloadl.c (getpayloadl): Use
_Float128 instead of long double.
* sysdeps/ieee754/ldbl-128/s_setpayloadl_main.c (FUNC): Likewise.
