Paul Eggert [Wed, 19 Sep 2018 20:16:14 +0000 (13:16 -0700)]
Fix mktime localtime offset confusion
[BZ #23603]
* include/time.h (__mktime_internal): The localtime offset is now
of type long int instead of time_t. This is the longstanding type
in glibc, and it is more than enough to represent difference
between localtime and gmtime even if it is 32 bits and time_t is
64. Changing it now will let us avoid an unnecessary change when
time_t is widened to 64 bits on 32-bit platforms.
* time/mktime-internal.h (mktime_offset_t): Now long int.
Paul Eggert [Wed, 19 Sep 2018 20:16:14 +0000 (13:16 -0700)]
Merge mktime, timegm from upstream Gnulib
[BZ #23603][BZ #16346]
This fixes some obscure problems with integer overflow.
Although it looks scary, it is almost all a byte-for-byte copy
from Gnulib, and the Gnulib code has been tested reasonably well.
* include/intprops.h: New file, copied from Gnulib.
* include/verify.h, time/mktime-internal.h:
New tiny files, simplified from Gnulib.
* time/mktime.c: Copy from Gnulib. This has the following changes:
Do not include config.h if DEBUG_MKTIME is nonzero.
Include stdbool.h, intprops.h, verify.h.
Include string.h only if needed.
Include stdlib.h on MS-Windows.
Include mktime-internal.h.
(DEBUG_MKTIME): Default to 0, and simplify later uses.
(NEED_MKTIME_INTERNAL, NEED_MKTIME_WINDOWS)
(NEED_MKTIME_WORKING): Give default values to pacify -Wundef,
which glibc uses. Default NEED_MKTIME_WORKING to DEBUG_MKTIME, to
simplify later conditionals; default the others to zero. Use
these conditionals to express only the code needed on the current
platform. In uses of these conditionals, explicitly spell out how
_LIBC affects things, so it’s easier to review from a glibc
viewpoint.
(WRAPV): Remove; no longer needed now that we have
systematic overflow checking.
(my_tzset, __tzset) [!_LIBC]: New function and macro, to better
compartmentalize tzset issues. Move system-dependent tzsettish
code here from mktime.
(verify): Remove; now done by verify.h. All uses changed.
(long_int): Use a more-conservative definition, to avoid
integer overflow.
(SHR): Remove, replacing with ...
(shr): New function, which means we needn’t worry about side
effects in args, and conversion analysis is simpler.
(TYPE_IS_INTEGER, TYPE_TWOS_COMPLEMENT, TYPE_SIGNED, TYPE_MINIMUM)
(TYPE_MAXIMUM, TIME_T_MIN, TIME_T_MAX, TIME_T_MIDPOINT)
(time_t_avg, time_t_add_ok): Remove.
(mktime_min, mktime_max): New constants.
(leapyear, isdst_differ): Use bool for booleans.
(ydhms_diff, guess_time_tm, ranged_convert, __mktime_internal):
Use long_int, not time_t, for mktime differences.
(long_int_avg): New function, replacing time_t_avg.
INT_ADD_WRAPV replaces time_t_add_ok.
(guess_time_tm): 6th arg is now long_int, not time_t const *.
All uses changed.
(convert_time): New function.
(ranged_convert): Use it.
(__mktime_internal): Last arg now points to mktime_offset_t, not
time_t. All uses changed. This is a no-op on glibc, where
mktime_offset_t is always time_t. Use int, not time_t, for UTC
offset guess. Directly check for integer overflow instead of
using a heuristic that works only 99.9...% of the time.
Access *OFFSET only once, to avoid an unlikely race if the
compiler delays a load and if this cascades into a signed integer
overflow.
(mktime): Move tzsettish code to my_tzset, and move
localtime_offset to within mktime so that it doesn’t
need a separate ifdef.
(main) [DEBUG_MKTIME]: Speed up by using localtime_r
instead of localtime.
* time/timegm.c: Copy from Gnulib. This has the following changes:
Include mktime-internal.h.
[!_LIBC]: Include config.h and time.h. Do not include
timegm.h or time_r.h. Make __mktime_internal a macro,
and include mktime-internal.h to get its declaration.
(timegm): Temporary is now mktime_offset_t, not time_t.
This affects only Gnulib.
The generic strstr in GLIBC 2.28 fails to match huge needles. The optimized
AVAILABLE macro reads ahead a large fixed amount to reduce the overhead of
repeatedly checking for the end of the string. However if the needle length
is larger than this, two_way_long_needle may confuse this as meaning the end
of the string and return NULL. This is fixed by adding the needle length to
the amount to read ahead.
Szabolcs Nagy [Wed, 13 Jun 2018 16:57:20 +0000 (17:57 +0100)]
Add new pow implementation
The algorithm is exp(y * log(x)), where log(x) is computed with about
1.3*2^-68 relative error (1.5*2^-68 without fma), returning the result
in two doubles, and the exp part uses the same algorithm (and lookup
tables) as exp, but takes the input as two doubles and a sign (to handle
negative bases with odd integer exponent). The __exp1 internal symbol
is no longer necessary.
There is separate code path when fma is not available but the worst case
error is about 0.54 ULP in both cases. The lookup table and consts for
log are 4168 bytes. The .rodata+.text is decreased by 37908 bytes on
aarch64. The non-nearest rounding error is less than 1 ULP.
Improvements on Cortex-A72 compared to current glibc master:
pow thruput: 2.40x in [0.01 11.1]x[0.01 11.1]
pow latency: 1.84x in [0.01 11.1]x[0.01 11.1]
Tested on
aarch64-linux-gnu (defined __FP_FAST_FMA, TOINT_INTRINSICS) and
arm-linux-gnueabihf (!defined __FP_FAST_FMA, !TOINT_INTRINSICS) and
x86_64-linux-gnu (!defined __FP_FAST_FMA, !TOINT_INTRINSICS) and
powerpc64le-linux-gnu (defined __FP_FAST_FMA, !TOINT_INTRINSICS) targets.
Joseph Myers [Tue, 18 Sep 2018 16:12:56 +0000 (16:12 +0000)]
Unify many bits/mman.h headers.
Many bits/mman.h headers for Linux architectures have exactly the same
contents, up to whitespace, comments and the number of leading 0s on
constants. Specifically, this applies to architectures that, in the
Linux kernel, either have no uapi/asm/mman.h, or have one that
includes asm-generic/mman.h without any changes or additions relevant
to glibc (this last case is the one that applies to Arm).
It's not useful to have to duplicate the set of MAP_* constants in
glibc for all such architectures and any new architectures with that
property. Thus, this patch creates a generic
sysdeps/unix/sysv/linux/bits/mman.h and removes all the
architecture-specific versions that become unnecessary.
Further unification remains possible after this patch. For example,
the new bits/mman.h could become bits/mman-map-flags-generic.h so that
it could also be used by architecture-specific bits/mman.h headers on
architectures that use the generic flags but add architecture-specific
ones to them. That would allow this common set of MAP_* definitions
to be used on ia64 and x86 as well (architectures that include
asm-generic/mman.h from their own uapi/asm/mman.h but define
additional MAP_* values of their own).
Joseph Myers [Tue, 18 Sep 2018 13:24:14 +0000 (13:24 +0000)]
Fix ldbl-128ibm ceill, floorl inlining of ceil, floor.
The ldbl-128ibm implementations of ceill and floorl call the
corresponding double functions. This patch fixes those
implementations to call those functions as ceil and floor rather than
as __ceil and __floor, so that the proper inlining takes place when
possible, while including local asm redirections for when the
functions are not inlined since NO_MATH_REDIRECT applies to the double
functions as well as to the long double ones.
Tested with build-many-glibcs.py for all its powerpc configurations.
* sysdeps/ieee754/ldbl-128ibm/s_ceill.c (ceil): Redirect to
__ceil.
(__ceill): Call ceil instead of __ceil.
* sysdeps/ieee754/ldbl-128ibm/s_floorl.c (floor): Redirect to
__floor.
(__floorl): Call floor instead of __floor.
Joseph Myers [Mon, 17 Sep 2018 20:42:06 +0000 (20:42 +0000)]
Use ceil functions not __ceil functions in glibc libm.
Continuing the move to use, within libm, public names for libm
functions that can be inlined as built-in functions on many
architectures, this patch moves calls to __ceil functions to call the
corresponding ceil names instead, with asm redirection to __ceil when
the calls are not inlined.
Joseph Myers [Mon, 17 Sep 2018 12:41:14 +0000 (12:41 +0000)]
Update siginfo constants from Linux kernel (bug 21286).
As of Linux 4.17, siginfo headers in the Linux kernel have been
largely unified across architectures (so various constants are defined
with common values in include/uapi/asm-generic/siginfo.h even if not
all architectures can generate those particular constants).
This patch makes glibc reflect that unification and the current set of
constants in that header as of Linux 4.18. Various constants are
added to bits/siginfo-consts.h (under the same feature test macro
conditions as the other constants with the same prefix), and removed
from the ia64 bits/siginfo-consts-arch.h where they were previously
there - this is not limited to constants added by the unification.
Nothing is done about macros that are defined in
include/uapi/asm-generic/siginfo.h with names with leading '__' (some
of those are ia64-specific ones that remain in the ia64
bits/siginfo-consts-arch.h without the leading '__' there).
A consequence of these changes is that TRAP_HWBKPT becomes available
on AArch64 and all other architectures as requested in bug 21286.
Tested for x86_64; tested with build-many-glibcs.py for ia64.
Joseph Myers [Fri, 14 Sep 2018 16:02:57 +0000 (16:02 +0000)]
Fix MIPS n32 pr_sigpend, pr_sighold, pr_flag type (bug 23656).
As discussed at
<https://sourceware.org/ml/libc-alpha/2018-09/msg00191.html> and
followup discussions, the MIPS n32 definitions of pr_sigpend and
pr_sighold in struct elf_prstatus, and pr_flag in struct elf_prpsinfo,
are wrong to use unsigned long long int; actual n32 core dumps use a
32-bit type there, so userspace unsigned long int is correct for all
MIPS ABIs. This patch removes the conditionals (also thereby aligning
the structures with other architectures and so facilitating future
unification of different versions of this header).
Tested with build-many-glibcs.py for its MIPS configurations.
Joseph Myers [Fri, 14 Sep 2018 13:21:33 +0000 (13:21 +0000)]
Fix sys/procfs.h pr_uid, pr_gid type (bug 23649).
As noted in
<https://sourceware.org/ml/libc-alpha/2018-09/msg00178.html>, glibc's
sys/procfs.h headers for microblaze, mips (n64), nios2 and riscv have
incorrect types for the pr_uid and pr_gid members of struct
elf_prpsinfo (as does the generic Linux version, but nothing uses
that).
This patch fixes those headers to use unsigned int. The generic Linux
version is also fixed, but I do *not* recommend making new
architectures use it yet. Rather, I think it should be reworked to
look more like a copy of the AArch64 version, but with a new
<bits/procfs.h> header included to provide register set definitions;
<bits/procfs.h> would then be architecture-specific while many
architectures could use the generic <sys/procfs.h>. This fix is
deliberately separate from any reworking to use a generic header more,
since it's possible there could be uses for backporting this fix but
not for backporting a subsequent cleanup.
Tested with build-many-glibcs.py. This of course doesn't provide much
validation of the structure layout; if the Linux kernel is fixed so
that "#include <linux/elfcore.h>" actually compiles with the headers
from "make headers_install" (and if the layout in both headers is
meant to be the same, whatever ABI we are building for), I have a test
that can be added to glibc to check the layout against that from the
Linux kernel.
[BZ #23649]
* sysdeps/unix/sysv/linux/microblaze/sys/procfs.h (struct
elf_prpsinfo): Use unsigned int for pr_uid and pr_gid.
* sysdeps/unix/sysv/linux/mips/sys/procfs.h (struct elf_prpsinfo):
Likewise.
* sysdeps/unix/sysv/linux/nios2/sys/procfs.h (struct
elf_prpsinfo): Likewise.
* sysdeps/unix/sysv/linux/riscv/sys/procfs.h (struct
elf_prpsinfo): Likewise.
* sysdeps/unix/sysv/linux/sys/procfs.h (struct elf_prpsinfo):
Likewise.
Joseph Myers [Fri, 14 Sep 2018 13:10:39 +0000 (13:10 +0000)]
Use rint functions not __rint functions in glibc libm.
Continuing the move to use, within libm, public names for libm
functions that can be inlined as built-in functions on many
architectures, this patch moves calls to __rint functions to call the
corresponding rint names instead, with asm redirection to __rint when
the calls are not inlined. The x86_64 math_private.h is removed as no
longer useful after this patch.
This patch is relative to a tree with my floor patch
<https://sourceware.org/ml/libc-alpha/2018-09/msg00148.html> applied,
and much the same considerations arise regarding possibly replacing an
IFUNC call with a direct inline expansion.
Joseph Myers [Fri, 14 Sep 2018 13:09:01 +0000 (13:09 +0000)]
Use floor functions not __floor functions in glibc libm.
Similar to the changes that were made to call sqrt functions directly
in glibc, instead of __ieee754_sqrt variants, so that the compiler
could inline them automatically without needing special inline
definitions in lots of math_private.h headers, this patch makes libm
code call floor functions directly instead of __floor variants,
removing the inlines / macros for x86_64 (SSE4.1) and powerpc
(POWER5).
The redirection used to ensure that __ieee754_sqrt does still get
called when the compiler doesn't inline a built-in function expansion
is refactored so it can be applied to other functions; the refactoring
is arranged so it's not limited to unary functions either (it would be
reasonable to use this mechanism for copysign - removing the inline in
math_private_calls.h but also eliminating unnecessary local PLT entry
use in the cases (powerpc soft-float and e500v1, for IBM long double)
where copysign calls don't get inlined).
The point of this change is that more architectures can get floor
calls inlined where they weren't previously (AArch64, for example),
without needing special inline definitions in their math_private.h,
and existing such definitions in math_private.h headers can be
removed.
Note that it's possible that in some cases an inline may be used where
an IFUNC call was previously used - this is the case on x86_64, for
example. I think the direct calls to floor are still appropriate; if
there's any significant performance cost from inline SSE2 floor
instead of an IFUNC call ending up with SSE4.1 floor, that indicates
that either the function should be doing something else that's faster
than using floor at all, or it should itself have IFUNC variants, or
that the compiler choice of inlining for generic tuning should change
to allow for the possibility that, by not inlining, an SSE4.1 IFUNC
might be called at runtime - but not that glibc should avoid calling
floor internally. (After all, all the same considerations would apply
to any user program calling floor, where it might either be inlined or
left as an out-of-line call allowing for a possible IFUNC.)
Joseph Myers [Wed, 12 Sep 2018 20:31:24 +0000 (20:31 +0000)]
Include most of elf/ modules-names in modules-names-tests.
I'm testing a patch to let the compiler expand calls to floor in libm
as built-in function calls as much as possible, instead of calling
__floor, so that no architecture-specific __floor inlines are needed,
and then to arrange for non-inlined calls to end up calling __floor,
as done with sqrt and __ieee754_sqrt.
This shows up elf/tst-relsort1mod2.c calling floor, which must not be
converted to a call to __floor. Now, while an IS_IN (libm)
conditional could be added to the existing conditionals on such
redirections in include/math.h, the _ISOMAC conditional ought to
suffice (code in other glibc libraries shouldn't be calling floor or
sqrt anyway, as they aren't provided in libc and the other libraries
don't link with libm). But while tests are mostly now built with
_ISOMAC defined, test modules in modules-names aren't unless also
listed in modules-names-tests.
As far as I can see, all the modules in modules-names in elf/ are in
fact parts of tests and so listing them in modules-names-tests is
appropriate, so they get built with something closer to the headers
used for user code, except in a few cases that actually rely on
something from internal headers. This patch duly sets
modules-names-tests there accordingly (filtering out those tests that
fail to build without internal headers).
Tested for x86_64, and with build-many-glibcs.py.
* elf/Makefile (modules-names-tests): New variable.
Szabolcs Nagy [Wed, 13 Jun 2018 16:48:52 +0000 (17:48 +0100)]
Add new log2 implementation
Similar algorithm is used as in log: log2(2^k x) = k + log2(c) + log2(x/c)
where the last term is approximated by a polynomial of x/c - 1, the first
order coefficient is about 1/ln2 in this case.
There is separate code path when fma instruction is not available for
computing x/c - 1 precisely, for which the table size is doubled.
The worst case error is 0.547 ULP (0.55 without fma), the read only
global data size is 1168 bytes (2192 without fma) on aarch64. The
non-nearest rounding error is less than 1 ULP.
Improvements on Cortex-A72 compared to current glibc master:
log2 thruput: 2.00x in [0.01 11.1]
log2 latency: 2.04x in [0.01 11.1]
log2 thruput: 2.17x in [0.999 1.001]
log2 latency: 2.88x in [0.999 1.001]
Szabolcs Nagy [Wed, 13 Jun 2018 16:40:19 +0000 (17:40 +0100)]
Add new log implementation
Optimized log using carefully generated lookup table with 1/c and log(c)
values for small intervalls around 1. The log(c) is very near a double
precision value, it has about 62 bits precision. The algorithm is
log(2^k x) = k log(2) + log(c) + log(x/c), where the last term is
approximated by a polynomial of x/c - 1. Near 1 a single polynomial of
x - 1 is used.
There is separate code path when fma instruction is not available for
computing x/c - 1 precisely, in which case the table size is doubled.
The code uses __builtin_fma under __FP_FAST_FMA to ensure it is inlined
as an instruction.
With the default configuration settings the worst case error is 0.519 ULP
(and 0.520 without fma), the rodata size is 2192 bytes (4240 without fma).
The non-nearest rounding error is less than 1 ULP.
Improvements on Cortex-A72 compared to current glibc master:
log thruput: 3.28x in [0.01 11.1]
log latency: 2.23x in [0.01 11.1]
log thruput: 1.56x in [0.999 1.001]
log latency: 1.57x in [0.999 1.001]
H.J. Lu [Wed, 12 Sep 2018 15:40:59 +0000 (08:40 -0700)]
i386: Use ENTRY and END in start.S [BZ #23606]
Wrapping the _start function with ENTRY and END to insert ENDBR32 at
function entry when CET is enabled. Since _start now includes CFI,
without "cfi_undefined (eip)", unwinder may not terminate at _start
and we will get
Program received signal SIGSEGV, Segmentation fault.
0xf7dc661e in ?? () from /lib/libgcc_s.so.1
Missing separate debuginfos, use: dnf debuginfo-install libgcc-8.2.1-3.0.fc28.i686
(gdb) bt
#0 0xf7dc661e in ?? () from /lib/libgcc_s.so.1
#1 0xf7dc7c18 in _Unwind_Backtrace () from /lib/libgcc_s.so.1
#2 0xf7f0d809 in __GI___backtrace (array=array@entry=0xffffc7d0,
size=size@entry=20) at ../sysdeps/i386/backtrace.c:127
#3 0x08049254 in compare (p1=p1@entry=0xffffcad0, p2=p2@entry=0xffffcad4)
at backtrace-tst.c:12
#4 0xf7e2a28c in msort_with_tmp (p=p@entry=0xffffca5c, b=b@entry=0xffffcad0,
n=n@entry=2) at msort.c:65
#5 0xf7e29f64 in msort_with_tmp (n=2, b=0xffffcad0, p=0xffffca5c)
at msort.c:53
#6 msort_with_tmp (p=p@entry=0xffffca5c, b=b@entry=0xffffcad0, n=n@entry=5)
at msort.c:53
#7 0xf7e29f64 in msort_with_tmp (n=5, b=0xffffcad0, p=0xffffca5c)
at msort.c:53
#8 msort_with_tmp (p=p@entry=0xffffca5c, b=b@entry=0xffffcad0, n=n@entry=10)
at msort.c:53
#9 0xf7e29f64 in msort_with_tmp (n=10, b=0xffffcad0, p=0xffffca5c)
at msort.c:53
#10 msort_with_tmp (p=p@entry=0xffffca5c, b=b@entry=0xffffcad0, n=n@entry=20)
at msort.c:53
#11 0xf7e2a5b6 in msort_with_tmp (n=20, b=0xffffcad0, p=0xffffca5c)
at msort.c:297
#12 __GI___qsort_r (b=b@entry=0xffffcad0, n=n@entry=20, s=s@entry=4,
cmp=cmp@entry=0x8049230 <compare>, arg=arg@entry=0x0) at msort.c:297
#13 0xf7e2a84d in __GI_qsort (b=b@entry=0xffffcad0, n=n@entry=20, s=s@entry=4,
cmp=cmp@entry=0x8049230 <compare>) at msort.c:308
#14 0x080490f6 in main (argc=2, argv=0xffffcbd4) at backtrace-tst.c:39
FAIL: debug/backtrace-tst
[BZ #23606]
* sysdeps/i386/start.S: Include <sysdep.h>
(_start): Use ENTRY/END to insert ENDBR32 at entry when CET is
enabled. Add cfi_undefined (eip).
Joseph Myers [Tue, 11 Sep 2018 14:51:40 +0000 (14:51 +0000)]
Remove x86_64 math_private.h asms.
The x86_64 math_private.h has asm versions of the macros to
reinterpret between floating-point and integer types.
This is the sort of thing we now strongly discourage; the expectation
in such cases, where the generic C code gives the compiler all the
information needed about the required semantics, is that you should
get the compiler to do the right thing for the generic C code rather
than writing an asm version.
Trivial tests showed GCC generates the expected single instructions
for reinterpretation from floating point to integer. In the other
direction, it goes via memory when the asms don't; I asked about this
in GCC bug 87236 and was advised this was deliberate for generic
tuning because it was faster that way on some AMD processors (but
-mtune=intel, and -Os with the latest GCC, avoid going via memory).
The asms don't and can't know about those tuning details, so that's
evidence that they are actually making the code worse.
This patch removes the asms accordingly. Tested for x86_64.
Joseph Myers [Tue, 11 Sep 2018 12:02:28 +0000 (12:02 +0000)]
Add build-many-glibcs.py support for building more GCC libraries.
Every so often we get libsanitizer or libgo builds breaking with new
glibc because of some change in the glibc headers.
glibc's build-many-glibcs.py deliberately disables libsanitizer and
GCC languages other than C and C++ because the point is to test glibc
and find glibc problems (including problems shown up by new compiler
warnings in new GCC), not to test libsanitizer or libgo; if the
compiler build fails because of libsanitizer or libgo failing to
build, that could hide the existence of new problems in glibc.
However, it seems reasonable to have a non-default mode where
build-many-glibcs.py does build those additional pieces, which this
patch adds.
Note that I do not intend to run a build-many-glibcs.py bot with this
new option. If people concerned with libsanitizer, libgo or other
potentially affected GCC libraries wish to find out about such
problems more quickly, they may wish to run such a bot or bots (and to
monitor the results and fix issues found - obviously there will be
some overlap with issues found by my bots not using that option).
Note also that building a non-native Ada compiler requires a
sufficiently recent native (or build-x-host, in general) Ada compiler
to be used, possibly more or less the same version as being built.
That needs to be in the PATH when build-many-glibcs.py --full-gcc is
run; the script does not deal with setting up such a compiler (or any
of the other host tools needed for building GCC and glibc, beyond the
GMP / MPFR / MPC libraries), but perhaps it should, to avoid the need
to keep updating such a compiler manually when running a bot.
Tested by running build-many-glibcs.py with the new option, with
mainline GCC. There are build failures for various configurations,
which may be of interest to Go / Ada people even if you're not
interested in running such a bot:
* mips64 / mips64el (all configuration): ICE building libstdc++, as
seen without using the new option
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87156>.
* aarch64_be: error building libgo (little-endian aarch64 works fine):
version.go:67:13: error: expected ';' or ')' or newline
67 | BigEndian =
| ^
version.go:67:3: error: reference to undefined name 'BigEndian'
67 | BigEndian =
| ^
* arm (all configurations): error building libgo:
/scratch/jmyers/glibc/many9/src/gcc/libgo/go/internal/syscall/unix/getrandom_linux.go:29:5: error: reference to undefined name 'randomTrap'
29 | if randomTrap == 0 {
| ^
/scratch/jmyers/glibc/many9/src/gcc/libgo/go/internal/syscall/unix/getrandom_linux.go:38:34: error: reference to undefined name 'randomTrap'
38 | r1, _, errno := syscall.Syscall(randomTrap,
| ^
What's happening there is, I think, that the arm*b*-*-* case in
libgo/configure.ac is wrongly matching arm-glibc-linux-gnueabi with
the 'b' in the vendor part, and then something else is failing to
handle GOARCH=armbe. Given that you can have configurations with
multilibs of both endiannesses, endianness should always be detected
by configure.ac, for all architectures, using a compile test of
whether __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__, not based on textual
matches to the host (= target at top-level) triplet.
* armeb (all configurations): error building libada (for some reason
the Arm libada configuration seems to do different things for EH for
big-endian, which makes no sense to me and doesn't actually work):
a-exexpr.adb:87:06: "System.Exceptions.Machine" is not a predefined library unit
a-exexpr.adb:87:06: "Ada.Exceptions (body)" depends on "Ada.Exceptions.Exception_Propagation (body)"
a-exexpr.adb:87:06: "Ada.Exceptions.Exception_Propagation (body)" depends on "System.Exceptions.Machine (spec)"
* hppa: error building libgo (same error as for aarch64_be).
* ia64: ICE building libgo. I've filed
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=87281> for this.
* m68k: ICE in the Go front end building libgo
<https://gcc.gnu.org/bugzilla/show_bug.cgi?id=84948>.
* microblaze, microblazeel, nios2, sh3, sh3eb: build failure in libada
for lack of a libada port to those systems (I'm not sure sh3 would
actually need anything different from sh4):
a-cbdlli.ads:38:14: violation of restriction "No_Finalization" at system.ads:47
* i686-gnu: build failure in libada, might be fixed by the patch
attached to <https://gcc.gnu.org/bugzilla/show_bug.cgi?id=81103>
(not tested):
terminals.c:1115:13: fatal error: termio.h: No such file or directory
* scripts/build-many-glibcs.py (Context.__init__): Add full_gcc
argument.
(Config.build_gcc): Use --disable-libsanitizer for first GCC
build, but not for second build if --full-gcc. Use
--enable-languages=all for second build if --full-gcc.
(get_parser): Add --full-gcc option.
(main): Update call to Context.
Italian and Swiss locales: Use the correct separators (bug 10797).
CLDR and many other sources say that it_IT (Italian) should use a dot
(".") as a thousands separator and a comma (",") as a decimal separator.
For it_CH and de_CH CLDR says that they should use the Right Single
Quotation Mark ("’") as a thousands separator and a dot (".") as a
decimal separator. Consequently, the same rules are copied to all other
locales in Switzerland.
These rules apply to both LC_MONETARY and LC_NUMERIC.
[BZ #10797]
* localedata/locales/de_CH (mon_thousands_sep): Use "<U2019>" (Right
Single Quotation Mark).
(thousands_sep): Likewise.
* localedata/locales/it_CH (LC_NUMERIC): Use “copy "de_CH"”.
* localedata/locales/it_IT (thousands_sep): Use ".".
(grouping): Use "3;3".
We've had issues before with build failures (with new GCC) in code
only built with --enable-obsolete-rpc or --enable-obsolete-nsl not
being reported for a while because build-many-glibcs.py does not test
those configure options. This patch adds configurations (32-bit and
64-bit) using those options so that in future we can notice quickly if
they start failing to build.
Tested the new configurations do build with GCC 8.
* scripts/build-many-glibcs.py (Context.add_all_configs): Add
x86_64 and i686 configs using --enable-obsolete-rpc
--enable-obsolete-nsl.
Stefan Liebler [Thu, 6 Sep 2018 12:27:03 +0000 (14:27 +0200)]
Fix segfault in maybe_script_execute.
If glibc is built with gcc 8 and -march=z900,
the testcase posix/tst-spawn4-compat crashes with a segfault.
In function maybe_script_execute, the new_argv array is dynamically
initialized on stack with (argc + 1) elements.
The function wants to add _PATH_BSHELL as the first argument
and writes out of bounds of new_argv.
There is an off-by-one because maybe_script_execute fails to count
the terminating NULL when sizing new_argv.
ChangeLog:
* sysdeps/unix/sysv/linux/spawni.c (maybe_script_execute):
Increment size of new_argv by one.
Indian and similar locales: Set the correct date format (bug 17426).
This commit also fixes d_fmt in bn_BD which is identical to bn_IN,
in ne_NP which is identical to ne_IN (not supported by Glibc but supported
by CLDR), and in ta_LK which is identical to ta_IN.
For those locales which are supported by CLDR data is imported from
CLDR v33. For others it is copied from those locales which were identical
before this commit.
* localedata/locales/kok_IN (d_fmt): Use "%-d-%-m-%y".
* localedata/locales/ne_NP (d_fmt): Use "%y//%-m//%-d".
* localedata/locales/sa_IN (d_fmt): Use "%-d-%m-%y".
* localedata/locales/te_IN (d_fmt): Use "%d-%m-%y".
Szabolcs Nagy [Mon, 12 Feb 2018 18:16:03 +0000 (18:16 +0000)]
Add new exp and exp2 implementations
Optimized exp and exp2 implementations using a lookup table for
fractional powers of 2. There are several variants, see e_exp_data.c,
they can be selected by modifying math_config.h allowing different
tradeoffs.
The default selection should be acceptable as generic libm code.
Worst case error is 0.509 ULP for exp and 0.507 ULP for exp2, on
aarch64 the rodata size is 2160 bytes, shared between exp and exp2.
On aarch64 .text + .rodata size decreased by 24912 bytes.
The non-nearest rounding error is less than 1 ULP even on targets
without efficient round implementation (although the error rate is
higher in that case). Targets with single instruction, rounding mode
independent, to nearest integer rounding and conversion can use them
by setting TOINT_INTRINSICS and adding the necessary code to their
math_private.h.
The __exp1 code uses the same algorithm, so the error bound of pow
increased a bit.
New double precision error handling code was added following the
style of the single precision error handling code.
Improvements on Cortex-A72 compared to current glibc master:
exp thruput: 1.61x in [-9.9 9.9]
exp latency: 1.53x in [-9.9 9.9]
exp thruput: 1.13x in [0.5 1]
exp latency: 1.30x in [0.5 1]
exp2 thruput: 2.03x in [-9.9 9.9]
exp2 latency: 1.64x in [-9.9 9.9]
For small (< 1) inputs the current exp code uses a separate algorithm
so the speed up there is less.
Was tested on
aarch64-linux-gnu (TOINT_INTRINSICS, fma contraction) and
arm-linux-gnueabihf (!TOINT_INTRINSICS, no fma contraction) and
x86_64-linux-gnu (!TOINT_INTRINSICS, no fma contraction) and
powerpc64le-linux-gnu (!TOINT_INTRINSICS, fma contraction) targets,
only non-nearest rounding ulp errors increase and they are within
acceptable bounds (ulp updates are in separate patches).
* NEWS: Mention exp and exp2 improvements.
* math/Makefile (libm-support): Remove t_exp.
(type-double-routines): Add math_err and e_exp_data.
* sysdeps/aarch64/libm-test-ulps: Update.
* sysdeps/arm/libm-test-ulps: Update.
* sysdeps/i386/fpu/e_exp_data.c: New file.
* sysdeps/i386/fpu/math_err.c: New file.
* sysdeps/i386/fpu/t_exp.c: Remove.
* sysdeps/ia64/fpu/e_exp_data.c: New file.
* sysdeps/ia64/fpu/math_err.c: New file.
* sysdeps/ia64/fpu/t_exp.c: Remove.
* sysdeps/ieee754/dbl-64/e_exp.c: Rewrite.
* sysdeps/ieee754/dbl-64/e_exp2.c: Rewrite.
* sysdeps/ieee754/dbl-64/e_exp_data.c: New file.
* sysdeps/ieee754/dbl-64/e_pow.c (__ieee754_pow): Update error bound.
* sysdeps/ieee754/dbl-64/eexp.tbl: Remove.
* sysdeps/ieee754/dbl-64/math_config.h: New file.
* sysdeps/ieee754/dbl-64/math_err.c: New file.
* sysdeps/ieee754/dbl-64/t_exp.c: Remove.
* sysdeps/ieee754/dbl-64/t_exp2.h: Remove.
* sysdeps/ieee754/dbl-64/uexp.h: Remove.
* sysdeps/ieee754/dbl-64/uexp.tbl: Remove.
* sysdeps/m68k/m680x0/fpu/e_exp_data.c: New file.
* sysdeps/m68k/m680x0/fpu/math_err.c: New file.
* sysdeps/m68k/m680x0/fpu/t_exp.c: Remove.
* sysdeps/powerpc/fpu/libm-test-ulps: Update.
* sysdeps/x86_64/fpu/libm-test-ulps: Update.
Joseph Myers [Wed, 5 Sep 2018 12:42:51 +0000 (12:42 +0000)]
Remove alpha math_private.h.
After some math_private.h cleanups (in particulat math-barriers.h
being split out), the only thing left in the alpha math_private.h was
macro definitions of __isnan and __isnanf, apparently (based on the
comments) intended to avoid problems with inline definitions in other
math_private.h files. Those inline definitions were removed in commit fe8c2b33aed067282ecac0cc3fdff1feef88548b, and the alpha math_private.h
is no longer needed; this patch removes it.
Tested with build-many-glibcs.py that installed stripped shared
libraries for alpha are unchanged by the patch.
Joseph Myers [Wed, 5 Sep 2018 11:53:35 +0000 (11:53 +0000)]
Move float128 inlines from sysdeps/generic/math_private.h to include/math.h.
Continuing the cleanup of math_private.h, with a view to it becoming
the header for the APIs defined therein and not also a header with
inline variants of math.h APIs, this patch moves inline definitions of
__isinff128 and fabsf128 to include/math.h, so that any users of
math.h in glibc automatically get the optimized functions rather than
quietly missing them if they do not also include math_private.h.
Tested for x86_64 and x86, and with build-many-glibcs.py with GCC 6.
There are changes to installed stripped libc.so on configurations with
distinct _Float128, because of __printf_fp_l code that now gets the
__isinff128 inline where previously it called the out-of-line
function because of the lack of a math_private.h call. It seems
appropriate that this code does get the inline (as it would
automatically with GCC 7 and later when the built-in function is used)
rather than being the only place in glibc that does not.
* sysdeps/generic/math_private.h
[__HAVE_DISTINCT_FLOAT128 && !__GNUC_PREREQ (7, 0)] (__isinff128):
Move this inline function ....
[__HAVE_DISTINCT_FLOAT128] (fabsf128): And this one ....
* include/math.h [!_ISOMAC]: To here....
Joseph Myers [Tue, 4 Sep 2018 19:52:06 +0000 (19:52 +0000)]
Move fenv.h soft-float inlines from fenv_private.h to include/fenv.h.
<fenv_private.h> has inline versions of various <fenv.h> functions,
and their __fe* variants, for systems (generally soft-float) without
support for floating-point exceptions, rounding modes or both.
Having these inlines in a separate header introduces a risk of a
source file including <fenv.h> and compiling OK on x86_64, but failing
to compile (because the feraiseexcept inline is actually a macro that
discards its argument, to avoid the need for #ifdef FE_INVALID
conditionals), or not being properly optimized, on systems without the
exceptions and rounding modes support (when these inlines were in
math_private.h, we had a few cases where this broke the build because
there was no obvious reason for a file to need math_private.h and it
didn't need that header on x86_64). By moving those inlines to
include/fenv.h, this risk can be avoided, and fenv_private.h becomes
more clearly defined as specifically the header for the internal
libc_fe* and SET_RESTORE_ROUND* interfaces.
This patch makes that move, removing fenv_private.h includes that are
no longer needed (or replacing them by fenv.h includes in a few cases
that didn't already have such an include).
Tested for x86_64 and x86, and tested with build-many-glibcs.py that
installed stripped shared libraries are unchanged by the patch.
Joseph Myers [Mon, 3 Sep 2018 21:09:04 +0000 (21:09 +0000)]
Do not include fenv_private.h in math_private.h.
Continuing the clean-up related to the catch-all math_private.h
header, this patch stops math_private.h from including fenv_private.h.
Instead, fenv_private.h is included directly from those users of
math_private.h that also used interfaces from fenv_private.h. No
attempt is made to remove unused includes of math_private.h, but that
is a natural followup.
(However, since math_private.h sometimes defines optimized versions of
math.h interfaces or __* variants thereof, as well as defining its own
interfaces, I think it might make sense to get all those optimized
versions included from include/math.h, not requiring a separate header
at all, before eliminating unused math_private.h includes - that
avoids a file quietly becoming less-optimized if someone adds a call
to one of those interfaces without restoring a math_private.h include
to that file.)
There is still a pitfall that if code uses plain fe* and __fe*
interfaces, but only includes fenv.h and not fenv_private.h or (before
this patch) math_private.h, it will compile on platforms with
exceptions and rounding modes but not get the optimized versions (and
possibly not compile) on platforms without exception and rounding mode
support, so making it easy to break the build for such platforms
accidentally.
I think it would be most natural to move the inlines / macros for fe*
and __fe* in the case of no exceptions and rounding modes into
include/fenv.h, so that all code including fenv.h with _ISOMAC not
defined automatically gets them. Then fenv_private.h would be purely
the header for the libc_fe*, SET_RESTORE_ROUND etc. internal
interfaces and the risk of breaking the build on other platforms than
the one you tested on because of a missing fenv_private.h include
would be much reduced (and there would be some unused fenv_private.h
includes to remove along with unused math_private.h includes).
Tested for x86_64 and x86, and tested with build-many-glibcs.py that
installed stripped shared libraries are unchanged by this patch.
Joseph Myers [Fri, 31 Aug 2018 21:19:51 +0000 (21:19 +0000)]
Replace conform/list-header-symbols.pl with a Python script.
Continuing the move of test code from Perl to Python (which seems
uncontroversial, unlike dependencies on Python in the actual build of
glibc), this patch replaces conform/list-header-symbols.pl with a
Python script, as a first step in converting the conform/ tests.
(conform/glibcconform.py is an equivalent to GlibcConform.pm,
containing code that will be relevant to move than one of the conform/
scripts.)
Tested for x86_64, including verifying that the symbol lists generated
are identical to those generated by the Perl version.
* conform/glibcconform.py: New file.
* conform/list-header-symbols.py: Likewise.
* conform/list-header-symbols.pl: Remove.
* conform/Makefile (tests-special): Only add linknamespace tests
if [PYTHON].
($(linknamespace-symlists-tests)): Use list-header-symbols.py.
H.J. Lu [Fri, 31 Aug 2018 20:07:47 +0000 (13:07 -0700)]
test-container: Use xcopy_file_range for cross-device copy [BZ #23597]
copy_file_range can't be used to copy a file from glibc source directory
to glibc build directory since they may be on different filesystems.
This patch adds xcopy_file_range for cross-device copy.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
[BZ #23597]
* support/Makefile (libsupport-routines): Add
support_copy_file_range and xcopy_file_range.
* support/support.h: Include <sys/types.h>.
(support_copy_file_range): New prototype.
* support/support_copy_file_range.c: New file. Copied and
modified from io/copy_file_range-compat.c.
* support/test-container.c (copy_one_file): Call xcopy_file_rang
instead of copy_file_range.
* support/xcopy_file_range.c: New file.
* support/xunistd.h (xcopy_file_range): New prototype.
Carlos O'Donell [Thu, 30 Aug 2018 15:01:33 +0000 (11:01 -0400)]
Fix test failure with -DNDEBUG.
The elf/tst-dlopen-aout.c test uses asserts to verify properties of the
test execution. Instead of using assert it should use xpthread_create
and xpthread_join to catch errors starting the threads and fail the
test. This shows up in Fedora 28 when building for i686-pc-linux-gnu
and using gcc 8.1.1.
Tested on i686, and fixes a check failure with -DNDEBUG.
Signed-off-by: Carlos O'Donell <carlos@redhat.com>
Florian Weimer [Thu, 30 Aug 2018 12:14:53 +0000 (14:14 +0200)]
reallocarray: Declare under _DEFAULT_SOURCE
Initially, this function was restricted to _GNU_SOURCE, but experience
shows that compatibility with existing build systems is improved if we
declare it under _DEFAULT_SOURCE as well.
Stefan Liebler [Thu, 30 Aug 2018 06:44:32 +0000 (08:44 +0200)]
Test stdlib/test-bz22786 exits now with unsupported if malloc fails.
The test tries to allocate more than 2^31 bytes which will always fail on s390
as it has maximum 2^31bit of memory.
Before commit 6c3a8a9d868a8deddf0d6dcc785b6d120de90523, this test returned
unsupported if malloc fails. This patch re enables this behaviour.
Furthermore support_delete_temp_files() failed to remove the temp directory
in this case as it is not empty due to the created symlink.
Thus the creation of the symlink is moved behind malloc.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
ChangeLog:
* stdlib/test-bz22786.c (do_test): Return EXIT_UNSUPPORTED
if malloc fails.
Joseph Myers [Wed, 29 Aug 2018 20:01:57 +0000 (20:01 +0000)]
Make gen-libm-test.py treat plus_oflow and minus_oflow as non-finite.
When converting gen-libm-test to Python, in one place I noted a bug in
the old Perl version that I preserved in the Python version so that
the generated output files were the same with both versions, as such
comparisons help give confidence in the correctness of such a rewrite
of a script. Now that the conversion has been done, this patch fixes
that bug, by arranging for tests with plus_oflow or minus_oflow
results (manually written tests in libm-test-*.inc that have
overflowing results that thus depend on the rounding mode) to be
properly treated as having non-finite results, and thus not run for
the __FINITE_MATH_ONLY__ tests. (As the affected tests in fact did
pass for __FINITE_MATH_ONLY__ testing, this is just a matter of
logical correctness in the choice of which tests run for that case,
rather than fixing any actual test failures.)
Tested for x86_64.
* math/gen-libm-test.py (gen_test_args_res): Also treat plus_oflow
and minus_oflow as non-finite.
Joseph Myers [Tue, 28 Aug 2018 20:48:49 +0000 (20:48 +0000)]
Split fenv_private.h out of math_private.h more consistently.
On some architectures, the parts of math_private.h relating to the
floating-point environment are in a separate file fenv_private.h
included from math_private.h. As this is purely an
architecture-specific convention used by several architectures,
however, all such architectures still need their own math_private.h,
even if it has nothing to do beyond #include <fenv_private.h> and
peculiarity of including the i386 file directly instead of having a
shared file in sysdeps/x86.
This patch makes the fenv_private.h name an architecture-independent
convention in glibc. The include of fenv_private.h from
math_private.h becomes architecture-independent (until callers are
updated to include fenv_private.h directly so the include from
math_private.h is no longer needed). Some architecture math_private.h
headers are removed if no longer needed, or renamed to fenv_private.h
if all they define belongs in that header; architecture fenv_private.h
headers now do require #include_next <fenv_private.h>. The i386
fenv_private.h file moves to sysdeps/x86/fpu/ to reflect how it is
actually shared with x86_64. The generic math_private.h gets a new
include of <stdbool.h>, as needed for bool in some prototypes in that
header (previously that was indirectly included via include/fenv.h,
which now only gets included too late in math_private.h, after those
prototypes).
Tested for x86_64 and x86, and tested with build-many-glibcs.py that
installed stripped shared libraries are unchanged by the patch.
* sysdeps/aarch64/fpu/fenv_private.h: New file. Based on ....
* sysdeps/aarch64/fpu/math_private.h: ... this file. All contents
moved to fenv_private.h except for ...
(TOINT_INTRINSICS): Kept in math_private.h.
(roundtoint): Likewise.
(converttoint): Likewise.
* sysdeps/arm/fenv_private.h: Change multiple-include guard to
[ARM_FENV_PRIVATE_H]. Include next <fenv_private.h>.
* sysdeps/arm/math_private.h: Remove.
* sysdeps/generic/fenv_private.h: New file. Contents moved from
....
* sysdeps/generic/math_private.h: ... this file. Include
<stdbool.h>. Do not include <fenv.h> or <get-rounding-mode.h>.
Include <fenv_private.h>. Remove functions and macros moved to
fenv_private.h.
* sysdeps/i386/fpu/math_private.h: Remove.
* sysdeps/mips/math_private.h: Move to ....
* sysdeps/mips/fpu/fenv_private.h: ... here. Change
multiple-include guard to [MIPS_FENV_PRIVATE_H]. Remove
[__mips_hard_float] conditional. Include next <fenv_private.h>.
* sysdeps/powerpc/fpu/fenv_private.h: Change multiple-include
guard to [POWERPC_FENV_PRIVATE_H]. Include next <fenv_private.h>.
* sysdeps/powerpc/fpu/math_private.h: Do not include
<fenv_private.h>.
* sysdeps/riscv/rvf/math_private.h: Move to ....
* sysdeps/riscv/rvf/fenv_private.h: ... here. Change
multiple-include guard to [RISCV_FENV_PRIVATE_H]. Include next
<fenv_private.h>.
* sysdeps/sparc/fpu/fenv_private.h: Change multiple-include guard
to [SPARC_FENV_PRIVATE_H]. Include next <fenv_private.h>.
* sysdeps/sparc/fpu/math_private.h: Remove.
* sysdeps/i386/fpu/fenv_private.h: Move to ....
* sysdeps/x86/fpu/fenv_private.h: ... here. Change
multiple-include guard to [X86_FENV_PRIVATE_H]. Include next
<fenv_private.h>.
* sysdeps/x86_64/fpu/math_private.h: Do not include
<sysdeps/i386/fpu/fenv_private.h>.
As done in commit 284f42bc778e487dfd5dff5c01959f93b9e0c4f5, memcmp
can be used after memchr to avoid the initialization overhead of the
two-way algorithm for the first match. This has shown improvement
>40% for first match.
Joseph Myers [Mon, 27 Aug 2018 18:52:41 +0000 (18:52 +0000)]
Move EXCEPTION_SET_FORCES_TRAP out of math-tests.h.
Completing the move of macros out of math-tests.h to smaller headers
following typo-proof conventions instead of using #ifndef, this patch
moves the EXCEPTION_SET_FORCES_TRAP macro out to its own
math-tests-trap-force.h header.
Tested with build-many-glibcs.py.
* sysdeps/generic/math-tests-trap-force.h: New file.
* sysdeps/generic/math-tests.h: Include <math-tests-trap-force.h>.
(EXCEPTION_SET_FORCES_TRAP): Do not define here.
* sysdeps/powerpc/math-tests.h: Remove file.
* sysdeps/powerpc/fpu/math-tests-trap-force.h: New file.
Paul Eggert [Sun, 26 Aug 2018 03:34:34 +0000 (20:34 -0700)]
regex: fix uninitialized memory access
I introduced this bug into gnulib in commit 8335a4d6c7b4448cd0bcb6d0bebf1d456bcfdb17 dated 2006-04-10;
eventually it was merged into glibc. The bug was found by
project-repo <bugs@feusi.co> and reported here:
https://lists.gnu.org/r/sed-devel/2018-08/msg00017.html
Diagnosis and draft fix reported by Assaf Gordon here:
https://lists.gnu.org/r/bug-gnulib/2018-08/msg00071.html
https://lists.gnu.org/r/bug-gnulib/2018-08/msg00142.html
* posix/regex_internal.c (build_wcs_upper_buffer):
Fix bug when mbrtowc returns 0.
Joseph Myers [Fri, 24 Aug 2018 19:18:16 +0000 (19:18 +0000)]
Move EXCEPTION_ENABLE_SUPPORTED out of math-tests.h.
Continuing moving macros out of math-tests.h to smaller headers
following typo-proof conventions instead of using #ifndef, this patch
moves the EXCEPTION_ENABLE_SUPPORTED macro out to its own
math-tests-trap.h header.
Tested with build-many-glibcs.py.
* sysdeps/generic/math-tests-trap.h: New file.
* sysdeps/generic/math-tests.h: Include <math-tests-trap.h>.
(EXCEPTION_ENABLE_SUPPORTED): Do not define here.
* sysdeps/aarch64/math-tests.h: Remove file.
* sysdeps/arm/math-tests.h: Likewise.
* sysdeps/riscv/math-tests.h: Likewise.
* sysdeps/aarch64/math-tests-trap.h: New file.
* sysdeps/arm/math-tests-trap.h: Likewise.
* sysdeps/riscv/math-tests-trap.h: Likewise.
Wilco Dijkstra [Fri, 24 Aug 2018 14:31:25 +0000 (15:31 +0100)]
Remove unused math files
Remove empty files due to the sin/cos improvements: k_sinf.c, k_cosf.c,
k_cos.c, k_sin.c. After the tanf change s_rem_pio2f.c and k_rem_pio2f.c
(and the ia64, m68k and powerpc equivalents) are no longer used,
so remove them. All e_rem_pio2.c files were already empty or commented
out, so remove them too. Passes build-many-glibcs.
Joseph Myers [Thu, 23 Aug 2018 23:41:13 +0000 (23:41 +0000)]
Move EXCEPTION_TESTS_* out of math-tests.h
Continuing moving macros out of math-tests.h to smaller headers
following typo-proof conventions instead of using #ifndef, this patch
moves the EXCEPTION_TESTS_* macros for individual types out to their
own sysdeps header.
As with ROUNDING_TESTS_*, there is no need to define these macros if
FE_ALL_EXCEPT == 0 and the individual exception macros are undefined;
thus, math-tests-exceptions.h headers are only needed for soft-float
ARM and RISC-V, while the other cases that defined these macros do not
need to do so (and the associated math-tests.h headers are thus
removed without needing replacement by math-tests-exceptions.h
headers).
Tested with build-many-glibcs.py.
* sysdeps/generic/math-tests-exceptions.h: New file.
* sysdeps/generic/math-tests.h: Include <math-tests-exceptions.h>.
(EXCEPTION_TESTS_float): Do not define here.
(EXCEPTION_TESTS_double): Likewise.
(EXCEPTION_TESTS_long_double): Likewise.
(EXCEPTION_TESTS_float128): Likewise.
* sysdeps/arm/math-tests.h [__SOFTFP__] (EXCEPTION_TESTS_float):
Likewise.
[__SOFTFP__] (EXCEPTION_TESTS_double): Likewise.
[__SOFTFP__] (EXCEPTION_TESTS_long_double): Likewise.
* sysdeps/arm/nofpu/math-tests-exceptions.h: New file.
* sysdeps/m68k/coldfire/math-tests.h: Remove file.
* sysdeps/mips/math-tests.h: Likewise.
* sysdeps/nios2/math-tests.h: Likewise.
* sysdeps/riscv/math-tests.h [!__riscv_flen]
(EXCEPTION_TESTS_float): Do not define here.
[!__riscv_flen] (EXCEPTION_TESTS_double): Likewise.
[!__riscv_flen] (EXCEPTION_TESTS_long_double): Likewise.
* sysdeps/riscv/nofpu/math-tests-exceptions.h: New file.
Wilco Dijkstra [Thu, 23 Aug 2018 11:38:16 +0000 (12:38 +0100)]
Speedup tanf range reduction
Speedup tanf range reduction by using the new sincosf range
reduction algorithm. Overall code quality is improved due to
inlining, so there is a speedup even if no range reduction is
required.
Paul Eggert [Thu, 23 Aug 2018 00:46:32 +0000 (17:46 -0700)]
regex: port Gnulib code to z/OS POSIX environment
Problem reported by Arnold Robbins in:
https://lists.gnu.org/r/bug-gnulib/2018-08/msg00129.html
* posix/regex_internal.h (__iswalnum, __towlower, __towupper) [!_LIBC]:
Undef.
Joseph Myers [Wed, 22 Aug 2018 15:22:06 +0000 (15:22 +0000)]
Don't redefine ROUNDING_TESTS_* in math/test-*-vlen*.h.
This patch completes the move of ROUNDING_TESTS_* macros to typo-proof
conventions by stopping redefining them in test-*-vlen*.h. Instead,
libm-test-driver.c is made to check TEST_MATHVEC when setting
non-to-nearest rounding modes.
Tested for x86_64.
* math/test-double-vlen2.h: Don't include <math-tests-rounding.h>.
(ROUNDING_TESTS_double): Remove.
* math/test-double-vlen4.h: Don't include <math-tests-rounding.h>.
(ROUNDING_TESTS_double): Remove.
* math/test-double-vlen8.h: Don't include <math-tests-rounding.h>.
(ROUNDING_TESTS_double): Remove.
* math/test-float-vlen16.h: Don't include <math-tests-rounding.h>.
(ROUNDING_TESTS_float): Remove.
* math/test-float-vlen4.h: Don't include <math-tests-rounding.h>.
(ROUNDING_TESTS_float): Remove.
* math/test-float-vlen8.h: Don't include <math-tests-rounding.h>.
(ROUNDING_TESTS_float): Remove.
* math/libm-test-driver.c (IF_ROUND_INIT_FE_DOWNWARD): Check
!TEST_MATHVEC here.
(IF_ROUND_INIT_FE_TOWARDZERO): Likewise.
(IF_ROUND_INIT_FE_UPWARD): Likewise.
Joseph Myers [Wed, 22 Aug 2018 12:25:50 +0000 (12:25 +0000)]
Move ROUNDING_TESTS_* out of math-tests.h.
Continuing moving macros out of math-tests.h to smaller headers
following typo-proof conventions instead of using #ifndef, this patch
moves the ROUNDING_TESTS_* macros for individual types out to their
own sysdeps header.
In the soft-float case where FE_TONEAREST is the only rounding mode
macro defined, there is no need to define ROUNDING_TESTS_*; it is only
necessary when rounding modes macros are defined that may not be
supported at runtime. Thus, the ROUNDING_TESTS_* definitions for some
configurations are just removed, not moved to new
math-tests-rounding.h headers; the only architectures needing
math-tests-rounding.h are those where the macros are defined in
bits/fenv.h because of the possibility of a soft-float compilation
using a hard-float glibc with the same ABI (i.e., ARM and RISC-V).
The test-*-vlen*.h headers, by using #undef, do not yet follow
typo-proof conventions (but they no longer implicitly rely on being
included before math-tests.h, and this area can always be cleaned up
further in future).
Tested with build-many-glibcs.py.
* sysdeps/generic/math-tests-rounding.h: New file.
* sysdeps/generic/math-tests.h: Include <math-tests-rounding.h>.
(ROUNDING_TESTS_float): Do not define here.
(ROUNDING_TESTS_double): Likewise.
(ROUNDING_TESTS_long_double): Likewise.
(ROUNDING_TESTS_float128): Likewise.
* math/test-double-vlen2.h: Include <math-tests-rounding.h>.
(ROUNDING_TESTS_double): Undefine before defining.
* math/test-double-vlen4.h: Include <math-tests-rounding.h>.
(ROUNDING_TESTS_double): Undefine before defining.
* math/test-double-vlen8.h: Include <math-tests-rounding.h>.
(ROUNDING_TESTS_double): Undefine before defining.
* math/test-float-vlen16.h: Include <math-tests-rounding.h>.
(ROUNDING_TESTS_float): Undefine before defining.
* math/test-float-vlen4.h: Include <math-tests-rounding.h>.
(ROUNDING_TESTS_float): Undefine before defining.
* math/test-float-vlen8.h: Include <math-tests-rounding.h>.
(ROUNDING_TESTS_float): Undefine before defining.
* sysdeps/arm/nofpu/math-tests-rounding.h: New file.
* sysdeps/arm/math-tests.h [__SOFTFP__] (ROUNDING_TESTS_float): Do
not define here.
[__SOFTFP__] (ROUNDING_TESTS_double): Likewise.
[__SOFTFP__] (ROUNDING_TESTS_long_double): Likewise.
* sysdeps/riscv/nofpu/math-tests-rounding.h: New file.
* sysdeps/riscv/math-tests.h [!__riscv_flen]
(ROUNDING_TESTS_float): Do not define here.
[!__riscv_flen] (ROUNDING_TESTS_double): Likewise.
[!__risv_flen] (ROUNDING_TESTS_long_double): Likewise.
* sysdeps/m68k/coldfire/math-tests.h [!__mcffpu__]
(ROUNDING_TESTS_float): Likewise.
[!__mcffpu__] (ROUNDING_TESTS_double): Likewise.
[!__mcffpu__] (ROUNDING_TESTS_long_double): Likewise.
* sysdeps/mips/math-tests.h [__mips_soft_float]
(ROUNDING_TESTS_float): Likewise.
[__mips_soft_float] (ROUNDING_TESTS_double): Likewise.
[__mips_soft_float] (ROUNDING_TESTS_long_double): Likewise.
* sysdeps/nios2/math-tests.h (ROUNDING_TESTS_float): Likewise.
(ROUNDING_TESTS_double): Likewise.
(ROUNDING_TESTS_long_double): Likewise.
Joseph Myers [Tue, 21 Aug 2018 15:57:22 +0000 (15:57 +0000)]
Update netinet/tcp.h from Linux 4.18.
This patch adds constants from netinet/tcp.h in Linux 4.18, and an
associated struct tcp_zerocopy_receive, to sysdeps/gnu/netinet/tcp.h.
The new TCP_REPAIR_* constants seemed sufficiently related to those
already present to include them.
Note that this patch does not include additions to struct tcp_info;
there are many other elements in this structure in the Linux kernel
that are not included in the glibc version (which was last extended in
2007, it seems). Such additions to the end of the structure may be OK
with the expected way it is used (size passed explicitly to the kernel
with getsockopt), but in principle any change to the size of a type
provided by glibc is an ABI change for external applications /
libraries using that type in their ABIs, and has the associated risks
of such a change.
Tested for x86_64.
* sysdeps/gnu/netinet/tcp.h (TCP_ZEROCOPY_RECEIVE): New macro.
(TCP_INQ): Likewise.
(TCP_CM_INQ): Likewise.
(TCP_REPAIR_ON): Likewise.
(TCP_REPAIR_OFF): Likewise.
(TCP_REPAIR_OFF_NO_WP): Likewise.
(struct tcp_zerocopy_receive): New type.
Joseph Myers [Mon, 20 Aug 2018 15:48:16 +0000 (15:48 +0000)]
Update struct signalfd_siginfo from Linux 4.18.
This patch updates struct signalfd_siginfo in sys/signalfd.h with new
members from Linux 4.18 (plus ssi_addr_lsb, added to the kernel in
2.6.37 without being added to sys/signalfd.h at that time). The
__pad2 member name follows the kernel and the existing __pad name.
Florian Weimer [Mon, 20 Aug 2018 12:57:13 +0000 (14:57 +0200)]
malloc: Add ChangeLog for accidentally committed change
Commit b90ddd08f6dd688e651df9ee89ca3a69ff88cd0c ("malloc: Additional
checks for unsorted bin integrity I.") was committed without a
whitespace fix, so it is adjusted here as well.
powerpc: Remove powerpc specific sinf and cosf optimization
New generic optimization of sinf and cosf introduced by commit 599cf3976679e1b345307d9c02057f02aa95528f shows improvement
compared to powerpc specific assembly version. Hence removing
the powerpc assembly versions to make use of generic code.
Istvan Kurucsai [Tue, 16 Jan 2018 13:54:32 +0000 (14:54 +0100)]
malloc: Additional checks for unsorted bin integrity I.
On Thu, Jan 11, 2018 at 3:50 PM, Florian Weimer <fweimer@redhat.com> wrote:
> On 11/07/2017 04:27 PM, Istvan Kurucsai wrote:
>>
>> + next = chunk_at_offset (victim, size);
>
>
> For new code, we prefer declarations with initializers.
Noted.
>> + if (__glibc_unlikely (chunksize_nomask (victim) <= 2 * SIZE_SZ)
>> + || __glibc_unlikely (chunksize_nomask (victim) >
>> av->system_mem))
>> + malloc_printerr("malloc(): invalid size (unsorted)");
>> + if (__glibc_unlikely (chunksize_nomask (next) < 2 * SIZE_SZ)
>> + || __glibc_unlikely (chunksize_nomask (next) >
>> av->system_mem))
>> + malloc_printerr("malloc(): invalid next size (unsorted)");
>> + if (__glibc_unlikely ((prev_size (next) & ~(SIZE_BITS)) !=
>> size))
>> + malloc_printerr("malloc(): mismatching next->prev_size
>> (unsorted)");
>
>
> I think this check is redundant because prev_size (next) and chunksize
> (victim) are loaded from the same memory location.
I'm fairly certain that it compares mchunk_size of victim against
mchunk_prev_size of the next chunk, i.e. the size of victim in its
header and footer.
>> + if (__glibc_unlikely (bck->fd != victim)
>> + || __glibc_unlikely (victim->fd != unsorted_chunks (av)))
>> + malloc_printerr("malloc(): unsorted double linked list
>> corrupted");
>> + if (__glibc_unlikely (prev_inuse(next)))
>> + malloc_printerr("malloc(): invalid next->prev_inuse
>> (unsorted)");
>
>
> There's a missing space after malloc_printerr.
Noted.
> Why do you keep using chunksize_nomask? We never investigated why the
> original code uses it. It may have been an accident.
You are right, I don't think it makes a difference in these checks. So
the size local can be reused for the checks against victim. For next,
leaving it as such avoids the masking operation.
> Again, for non-main arenas, the checks against av->system_mem could be made
> tighter (against the heap size). Maybe you could put the condition into a
> separate inline function?
We could also do a chunk boundary check similar to what I proposed in
the thread for the first patch in the series to be even more strict.
I'll gladly try to implement either but believe that refining these
checks would bring less benefits than in the case of the top chunk.
Intra-arena or intra-heap overlaps would still be doable here with
unsorted chunks and I don't see any way to counter that besides more
generic measures like randomizing allocations and your metadata
encoding patches.
I've attached a revised version with the above comments incorporated
but without the refined checks.
Thanks,
Istvan
From a12d5d40fd7aed5fa10fc444dcb819947b72b315 Mon Sep 17 00:00:00 2001
From: Istvan Kurucsai <pistukem@gmail.com>
Date: Tue, 16 Jan 2018 14:48:16 +0100
Subject: [PATCH v2 1/1] malloc: Additional checks for unsorted bin integrity
I.
Ensure the following properties of chunks encountered during binning:
- victim chunk has reasonable size
- next chunk has reasonable size
- next->prev_size == victim->size
- valid double linked list
- PREV_INUSE of next chunk is unset
Pochang Chen [Thu, 16 Aug 2018 19:24:24 +0000 (15:24 -0400)]
malloc: Verify size of top chunk.
The House of Force is a well-known technique to exploit heap
overflow. In essence, this exploit takes three steps:
1. Overwrite the size of top chunk with very large value (e.g. -1).
2. Request x bytes from top chunk. As the size of top chunk
is corrupted, x can be arbitrarily large and top chunk will
still be offset by x.
3. The next allocation from top chunk will thus be controllable.
If we verify the size of top chunk at step 2, we can stop such attack.
This patch moves little endian specific POWER9 optimization files to
sysdeps/powerpc/powerpc64/le and creates POWER9 ifunc functions
only for little endian.
[aarch64] Add an ASIMD variant of strlen for falkor
This variant of strlen uses vector loads and operations to reduce the
size of the code and also eliminate the non-ascii fallback. This
works very well for falkor because of its two vector units and
efficient vector ops. In the best case it reduces latency of cases in
bench-strlen by 48%, with gains throughout the benchmark.
strlen-walk also sees uniform gains in the 5%-15% range.
Overall the routine appears to work better than the stock one for falkor
regardless of the benchmark, length of string or cache state.
The same cannot be said of a53 and a72 though. a53 performance was
greatly reduced and for a72 it was a bit of a mixed bag, slightly on the
negative side but I reckon it might be fast in some situations.
* sysdeps/aarch64/strlen.S (__strlen): Rename to STRLEN.
[!STRLEN](STRLEN): Set to __strlen.
* sysdeps/aarch64/multiarch/strlen.c: New file.
* sysdeps/aarch64/multiarch/strlen_generic.S: Likewise.
* sysdeps/aarch64/multiarch/strlen_asimd.S: Likewise.
* sysdeps/aarch64/multiarch/ifunc-impl-list.c
(__libc_ifunc_impl_list): Add strlen.
* sysdeps/aarch64/multiarch/Makefile (sysdep_routines): Add
strlen_generic and strlen_asimd.
Wilco Dijkstra [Wed, 15 Aug 2018 15:01:21 +0000 (16:01 +0100)]
Use generic sinf/cosf in lgammaf_r
The internal functions __kernel_sinf and __kernel_cosf are used only by
lgammaf_r. Removing the internal functions and using the generic sinf
and cosf is better overall. Benchmarking on Cortex-A72 shows the generic
sinf and cosf are 1.4x and 2.3x faster in the range |x| < PI/4, and 0.66x
and 1.1x for |x| < PI/2, so it should make lgammaf_r faster on average.
GLIBC regression tests pass on AArch64.
* sysdeps/ieee754/flt-32/e_lgammaf_r.c (sin_pif): Use __sinf/__cosf.
* sysdeps/ieee754/flt-32/k_cosf.c (__kernel_cosf): Remove all code.
* sysdeps/ieee754/flt-32/k_sinf.c (__kernel_sinf): Likewise.
Wilco Dijkstra [Tue, 14 Aug 2018 09:45:59 +0000 (10:45 +0100)]
Improve performance of sinf and cosf
The second patch improves performance of sinf and cosf using the same
algorithms and polynomials. The returned values are identical to sincosf
for the same input. ULP definitions for AArch64 and x64 are updated.
Joseph Myers [Mon, 13 Aug 2018 21:35:27 +0000 (21:35 +0000)]
Update syscall-names.list for Linux 4.18.
This patch updates sysdeps/unix/sysv/linux/syscall-names.list for
Linux 4.18. The io_pgetevents and rseq syscalls are added to the
kernel on various architectures, so need to be mentioned in this file.
Tested with build-many-glibcs.py.
* sysdeps/unix/sysv/linux/syscall-names.list: Update kernel
version to 4.18.
(io_pgetevents): New syscall.
(rseq): Likewise.
Joseph Myers [Mon, 13 Aug 2018 17:20:46 +0000 (17:20 +0000)]
Update install.texi documentation of uses of Perl and Python.
The install.texi documentation of uses of Perl and Python is
substantially out of date.
The description of Perl is "to test the installation" (which I
interpret as referring to test-installation.pl), but it's used for
more tests than that, and to build the manual, and to regenerate one
file in the source tree.
The description of Python is only for pretty-printer tests, but it's
used for other tests / benchmarks as well (and for other internal uses
such as updating Unicode data, for which we already require Python 3,
but I think install.texi only needs to describe uses from the main
glibc Makefiles).
This patch updates the descriptions of what those tools are used for.
The Python information (and information about other tools for testing
pretty printers) was awkwardly in the middle of the general
description of building and testing glibc, rather than with the rest
of information about tools used in glibc build and test; this patch
moves the information about those tools into the main list.
Tested with regeneration of INSTALL as well as "make info" and "make
pdf".
* manual/install.texi (Configuring and compiling): Do not list
tools used for testing pretty printers here.
(Tools for Compilation): List Python, PExpect and GDB here.
Update descriptions of uses of Perl and Python.
* INSTALL: Regenerate.
Florian Weimer [Mon, 13 Aug 2018 12:05:00 +0000 (14:05 +0200)]
mbstowcs: Remove outdated comment
ISO C requires that there is no effect on any global conversion state,
so the change in commit 9f097308c7465443765d1e25699a4cf33eff5455 was
correct in princple.
Paul Eggert [Fri, 10 Aug 2018 21:19:05 +0000 (14:19 -0700)]
regex: Gnulib unibyte RRI uses bytes not chars
Adjust the non-glibc code to agree with what Gawk needs for
rational range interpretation (RRI) for regular expression ranges.
In unibyte locales, Gawk wants ranges to use the underlying byte
rather than the character code point. This change does not affect
glibc proper.
* posix/regcomp.c (parse_byte) [!LIBC && RE_ENABLE_I18N]:
In unibyte locales, use the byte value rather than
running it through btowc.
Joseph Myers [Fri, 10 Aug 2018 19:22:01 +0000 (19:22 +0000)]
Move SNAN_TESTS_* out of math-tests.h.
Continuing moving macros out of math-tests.h to smaller headers
following typo-proof conventions instead of using #ifndef, this patch
moves the SNAN_TESTS_* macros for individual types out to their own
sysdeps header (while the type-generic SNAN_TESTS wrapper for those
macros remains in math-tests.h).
Tested for x86_64 and x86, and with build-many-glibcs.py.
* sysdeps/generic/math-tests-snan.h: New file.
* sysdeps/generic/math-tests.h: Include <math-tests-snan.h>.
(SNAN_TESTS_float): Do not define here.
(SNAN_TESTS_double): Likewise.
(SNAN_TESTS_long_double): Likewise.
(SNAN_TESTS_float128): Likewise.
* sysdeps/i386/fpu/math-tests-snan.h: New file.
* sysdeps/i386/fpu/math-tests.h: Remove file.
* sysdeps/ia64/math-tests-snan.h: New file.
* sysdeps/ia64/math-tests.h: Remove file.
* sysdeps/x86/math-tests.h: Likewise.
* sysdeps/x86_64/fpu/math-tests-snan.h: New file.
Wilco Dijkstra [Fri, 10 Aug 2018 16:31:30 +0000 (17:31 +0100)]
Improve performance of sincosf
This patch is a complete rewrite of sincosf. The new version is
significantly faster, as well as simple and accurate.
The worst-case ULP is 0.5607, maximum relative error is 0.5303 * 2^-23 over
all 4 billion inputs. In non-nearest rounding modes the error is 1ULP.
The algorithm uses 3 main cases: small inputs which don't need argument
reduction, small inputs which need a simple range reduction and large inputs
requiring complex range reduction. The code uses approximate integer
comparisons to quickly decide between these cases.
The small range reducer uses a single reduction step to handle values up to
120.0. It is fastest on targets which support inlined round instructions.
The large range reducer uses integer arithmetic for simplicity. It does a
32x96 bit multiply to compute a 64-bit modulo result. This is more than
accurate enough to handle the worst-case cancellation for values close to
an integer multiple of PI/4. It could be further optimized, however it is
already much faster than necessary.
projects / thirdparty / glibc.git / log
