[PEI] Don't zero out noreg operands
A tail call may have $noreg operands.
Fixes a crash.
Reviewed By: xgupta
Differential Revision: https://reviews.llvm.org/D156485
This should fix an assertion failure building qemu, specifically those
parts using -fzero-call-used-regs.
Reported by: Daniel Berrangé <dan-freebsd@berrange.com>
PR: 277474
MFC after: 3 days
hostap MLME uses Linux data structures and definitions not available
in FreeBSD. The ability for hostapd to select the frequency (channel)
depends Linux MLME, though strictly it's not required. Work around the
Linux MLME requirement to configure device frequency.
The detailed description is: hostapd will only set the channel (frequency)
when Linux MLME is configured. Enabling NEED_AP_MLME will result in
numerous build errors due do Linux data structures and definitions not
available under FreeBSD. The code to set the frequency from the selected
channel is only within the NEED_AP_MLME code path because without MLME,
hostapd_get_hw_features() is an inline that always returns -1 whereas with
MLME hostapd_get_hw_features() will obtain hardware features from the
kernel. Until such time we simply set the frequency as configured.
PR: 276375
MFC after: 1 month
This brings in Unicode support, CSV support and a number of bug fixes.
They are described in _The AWK Programming Language_, Second Edition, by
Al Aho, Brian Kernighan, and Peter Weinberger (Addison-Wesley, 2024,
ISBN-13 978-0138269722, ISBN-10 0138269726).
Sponsored by: Netflix
[sanitizer][msan] fix AArch64 vararg support for KMSAN (#70660)
Cast StackSaveAreaPtr, GrRegSaveAreaPtr, VrRegSaveAreaPtr to pointers to
fix assertions in getShadowOriginPtrKernel().
Fixes: https://github.com/llvm/llvm-project/issues/69738
Patch by Mark Johnston.
Requested by: markj
MFC after: 3 days
traceroute hasn't had a vendor import since 2002, while since then it's
had several significant FreeBSD-specific commits. Since it's unlikely
another vendor import will happen, and to make the merge of traceroute6
into traceroute easier, import traceroute into usr.sbin.
Reviewed by: imp
Pull Request: https://github.com/freebsd/freebsd-src/pull/1100
create_qp_handle_resp_common_cleanup should be void
__ibv_cleanup_wq should use wq->cond for cond destroy
both issues were overlooked in:
a687910 ('Cleanup pthread locks in ofed RDMA verbs')
Signed-off-by: Bartosz Sobczak <bartosz.sobczak@intel.com>
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: sean.lim@dell.com, vangyzen@, erj@
MFC after: 1 day
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D43491
Silence the warning:
argument to 'sizeof' in 'strncpy' call is the same expression as the
source; did you mean to use the size of the destination?"
by using strlcpy() and the size of the destination (device).
No functional change intended.
Reviewed by: imp
Pull Request: https://github.com/freebsd/freebsd-src/pull/901
[libc++] Fix filesystem::remove_all() on FreeBSD (#79540)
remove_all_impl() opens the target path with O_NOFOLLOW, which fails if
the target is a symbolic link. On FreeBSD, rather than returning ELOOP,
openat() returns EMLINK. This is unlikely to change for compatibility
reasons, see https://bugs.freebsd.org/bugzilla/show_bug.cgi?id=214633 .
Thus, check for EMLINK as well.
Reported by: markj
PR: 276632
MFC after: 3 days
[compiler-rt] remove hexdump interception. (#79378)
a freebsd dev member reported a symbol conflict and intercepting this
had little value anyway.
This is one part of fixing the WITH_ASAN build. Some executables in the
base system define their own hexdump() symbol, which would otherwise
conflict with the ASan-intercepted one.
Reported by: markj
PR: 276597
MFC after: 3 days
This update fixes a bug that line breaks in printed numbers may not
match the line length set by the user. The value is printed correctly,
just not split as specified in some situations.
(cherry picked from commit 52a5ec1b178fd07651446c7e31b1512794a04dbf)
This update improves the implementation of the power function p() and
adds 3 new functions to the extended math library: min(), max(), and
i2rand().
(cherry picked from commit 0b4a06ab29a0da80f6cb5c99189054cb8e2f756c)
Introduce pfctl_add_rule_h(), which takes a pfctl_handle rather than a
file descriptor (which it didn't use). This means that library users can
open the handle while they're running as root, but later drop privileges
and still add rules to pf.
Sponsored by: Rubicon Communications, LLC ("Netgate")
Explicit Congestion Notification (ECN) is a mechanism that allows
end-to-end notification of network congestion without dropping packets
by explicitly setting the ECN code point (2 bits).
Per RFC 8087, section 3.5, network devices should not be configured to
change the ECN code point in the packets that they forward, except to
set the CE (Congestion Experienced) code point ('11') to signal
incipient congestion.
The current commit adds an -E flag to traceroute that crafts a packet
with an ECT(1) code point ('01').
If the packet is received back with a zero ECN code point ('00'), it
outputs that the hop in question erases or "bleaches" the ECN code point
values. Bleaching may occur for various reasons (including normalizing
packets to hide which equipment supports ECN). This policy prevents the
use of ECN by applications.
If the packet is received back with an all-ones ECN code point ('11'),
it outputs that the hop in question is experiencing "congestion".
If the packet is received back with a different ECN code point ('10'),
it outputs that the hop in question changes or "mangles" the ECN code
point values.
If the packet is received with the same ECN code point that was sent
('01'), it outputs that the hop has "passed" the ECN bits appropriately.
Inspired by: Darwin
Reviewed by: imp, markj
MFC after: 1 month
Pull Request: https://github.com/freebsd/freebsd-src/pull/879
Signal blocking originally showed up in 51c2afc4b65b2782 ("Support:
Don't call close again if we get EINTR"), but it was overzealous --
there are systems where the error is known to be fine.
This commit elides signal blocking for said systems (the list is
incomplete though).
Note close() can still fail for other reasons (like ENOSPC), in which
case an error will be returned while the fd slot is cleared up.
Reviewed by: dim
Differential Revision: https://reviews.freebsd.org/D42984
In commit 88640c0e8b the new EINTEGRITY errno value was added, and
this caused us to carry a patch for upstream libc++ since that time.
Because it can cause merge conflicts when importing libc++ code from
upstream, I have submitted an upstream pull request to get most of that
patch integrated.
It turns out that we do not need the errno.h part of it at all, since
all supported FreeBSD versions define EOWNERDEAD and ENOTRECOVERABLE,
and therefore the block that juggles with ELAST values is never used in
FreeBSD. At the moment it only applies to older versions of Linux, or
possibly other platforms.
Therefore the only part that needs to stay is the definition of a enum
errc value for EINTEGRITY, and this is made optional upon EINTEGRITY
being defined, to make it suitable for upstreaming.
No functional change is intended.
MFC after: 1 week
Building telnet with clang 18 results in the following warning:
contrib/telnet/telnet/telnet.c:231:5: error: 'snprintf' will always be truncated; specified size is 10, but format string expands to at least 11 [-Werror,-Wformat-truncation]
231 | snprintf(temp2, sizeof(temp2), "%c%c%c%c....%c%c", IAC, SB, TELOPT_COMPORT,
| ^
The temp2 buffer is 10 chars, while the format string also consists of
10 chars. Therefore, snprintf(3) will truncate the last character, 'SE'
(end sub negotation) in this case.
Bump the buffer to 11 chars to avoid truncation.
MFC after: 3 days
It appears that Clang 16 errors by default for the mismatched function
pointer types that are triggered by the fts callback (since glibc has a
callback type without the second const qualifier).
Fortunately, there is already code to handle glibc inside mandoc, we
just have to edit the checked-in config.h.
Reviewed By: imp, emaste
Differential Revision: https://reviews.freebsd.org/D42999
Before this commit, we only had the capability to check if a specific
capability was set (using cap_rights_is_set function). However, there
was no efficient method to determine if a cap_rights_t structure doesn't
contain any capability. The cap_rights_is_empty function addresses
this gap.
PR: 275330
Reported by: vini.ipsmaker@gmail.com
Reviewed by: emaste, markj
Differential Revision: https://reviews.freebsd.org/D42780
[Clang] Do not change the type of captured vars when checking lambda constraints
When checking the constraint of a lambda, we need to respect the constness
of the call operator when establishing the type of capture variables.
In D124351, this was done by adding const to the captured variable...
However, that would change the type of the variable outside of the scope
of the lambda, which is clearly not the desired outcome.
Instead, to ensure const-correctness, we need to populate
a LambdaScopeInfo with the capture variables before checking the
constraints of a generic lambda.
There is no changelog as I'd like to tentatively propose we backport
this change to RC3 as it is a regression introduced in the Clang 17
cycle.
Fixes#61267
Reviewed By: aaron.ballman, #clang-language-wg
Differential Revision: https://reviews.llvm.org/D158433
Merge commit 3ed9e9e3ace6 from llvm git (by Corentin Jabot):
[Clang] Add captures to the instantiation scope of lambda call operators
Like concepts checking, a trailing return type of a lambda
in a dependent context may refer to captures in which case
they may need to be rebuilt, so the map of local decl
should include captures.
This patch reveal a pre-existing issue.
`this` is always recomputed by TreeTransform.
`*this` (like all captures) only become `const`
after the parameter list.
However, if try to recompute the value of `this` (in a parameter)
during template instantiation while determining the type of the call operator,
we will determine it to be const (unless the lambda is mutable).
There is no good way to know at that point that we are in a parameter
or not, the easiest/best solution is to transform the type of this.
Note that doing so break a handful of HLSL tests.
So this is a prototype at this point.
Fixes#65067Fixes#63675
Reviewed By: erichkeane
Differential Revision: https://reviews.llvm.org/D159126
This fixes 'Assertion failed: (isa<LabelDecl>(D) && "declaration not
instantiated in this scope"), function findInstantiationOf' error when
building databases/mongodb44.
PR: 273753
MFC after: 1 month
This updates llvm, clang, compiler-rt, libc++, libunwind, lld, lldb and
openmp to llvm-project main llvmorg-17-init-19304-gd0b54bb50e51, the
last commit before the upstream release/17.x branch was created.
PR: 273753
MFC after: 1 month
In order to avoid running into the default net.local.dgram.maxdgram
of 2K currently when calling sendto(2) try to set the sndbuf size to
the maximum ctrl message size.
While on 14 and 15 this does not actually raise the limit anymore (and
be7c095ac9 raised it for syslogd and this),
FreeBSD 13 still requires this change and it will work as expected there.
In addition we always ensure a large enough send buffer this way
independent of kernel defaults.
The problem occured, e.g., when the scan_list result had enough BSSIDs
so the text output would exceed 2048 bytes.
Sponsored by: The FreeBSD Foundation
MFC after: 3 days
PR: 274990
Reviewed by: cy, adrian (with previous comment)
Differential Revision: https://reviews.freebsd.org/D42558
After 930a7c2ac6 ("compiler-rt: re-exec with ASLR disabled when
necessary") and 96fe7c8ab0 ("compiler-rt: support ReExec() on
FreeBSD"), binaries linked against the sanitizer libraries may segfault
due to procctl(2) being intercepted. Instead, the non-intercepted
internal_procctl() should be called.
Similarly, the ReExec() function that re-executes the binary after
turning off ASLR should not call elf_aux_info(3) and realpath(3), since
these will also be intercepted. Instead, loop directly over the elf aux
info vector to find the executable path, and avoid calling realpath(3)
since it is actually unwanted for this use case.
Fixes: 930a7c2ac6, 96fe7c8ab0
MFC after: 3 days
These are local additions that no longer make sense with the transition
to git.
This partially reverts a10f530f93.
Reviewed by: kp, imp
Differential Revision: https://reviews.freebsd.org/D42687
I added them per then-current practice when I imported them, but now
they are just gratuious local diffs.
Reviewed by: imp
Differential Revision: https://reviews.freebsd.org/D42686
We don't need to compile on a K&R compiler (and we've long ago lost the
ability to do so). It's not even clear if it ever worked with a pure K&R
compiler, but maybe it once did...
Sponsored by: Netflix
The pre-push testing I did turned out to be testing the old version with
the old testsuite (for reasons I don't understnad). There's issues with
the new version, the new test in the suite or (likely) both. Revert
until they can be chased down.
This should also fix the github CI that's gone red since this commit.
This reverts commit 3fd60a6b73, reversing
changes made to 194df014fe.
Sponsored by: Netflix
Merge in the November 2nd, 2023 version of one true awk.
This brings in Unicode support, CSV support and a number of bug fixes.
Sponsored by: Netflix
Reviewed by: delphij
Differential Revision: https://reviews.freebsd.org/D42447
Upstream tarball included config.h~.
This was previously manually removed from vendor where in fact it
should not have been as it would have caused a merge conflict.
Reported by: "Herbert J. Skuhra" <herbert@gojira.at>
MFC: 2 weeks
When print-ip-demux.c was introduced on ee67461e, the pfsync_ip_print
function was missed, causing tcpdump to treat pfsync packets on network
interfaces as an unknown protocol.
MFC after: 1 week
Sponsored by: InnoGames GmbH
Differential Revision: https://reviews.freebsd.org/D42504
Upstream is now https://github.com/zoulasc/blocklist/. Rename the
contrib directory and update Makefiles to match, in advance of the next
vendor branch update.
Sponsored by: The FreeBSD Foundation
I would like to use this file as an example of the FREEBSD-upgrade
convention, see D42302. libarchive is picked somewhat arbitrarily as a
longstanding piece of contrib software in FreeBSD.
- Remove SVN references (HEAD/trunk)
- Mention the vendor/libarchive git branch
- Update link to import instructions
- Remove $FreeBSD$
Reviewed by: mm, imp, emaste
Sponsored by: The FreeBSD Foundation
Differential Revision: https://reviews.freebsd.org/D42308
From upstream 0ee44ef3:
- Fix send of udp retries when ENOBUFS is returned. It stops looping
and also waits for the condition to go away. Reported by Florian
Obser.
PR: 274352
MFC after: 3 days
Merge commit '292d51198aa319c58f534549851e9c28486abdf4'
Trimming of the line feed is no longer necessary after
d993c6b0db
Currently the tests:
- bcachefs2
- gpkg-1-zst
- multiple
are failing, but a fix will be committed upstream.
This also reverts c5e957ad4 "file: fix test case for gpkg by removing the extra \n."
MFC after: 3 days
despite being ignored in our main .gitignore, such files can still be
merged from vendor branches.
Reviewed by: cy
Fixes: 8f76bb7dad (unbound: Vendor import 1.18.0)
MFC after: 2 weeks
X-MFC with: 8f76bb7dad
Differential Revision: https://reviews.freebsd.org/D41920
Some highlights from NEWS entries:
** Improved OpenSSL 3.0 compatibility.
** Support for hidraw(4) on FreeBSD; gh#597.
** Improved support for FIDO 2.1 authenticators.
PR: 273596
Relnotes: Yes
Sponsored by: The FreeBSD Foundation
On FreeBSD, pthread mutex, cond, and spinlocks allocate memory. On
Linux-based systems, these calls do not allocate memory. So there was a
safe assumption that the ofed RDMA verb calls do not need to explicitly
destroy the pthread locks. This assumption is false on FreeBSD. So let
us rearrange the code to cleanup the pthread locks.
Reviewed by: delphij
MFC after: 2 weeks
Sponsored by: Dell EMC Isilon
Differential Revision: https://reviews.freebsd.org/D41105
An explicit tzset() call is usually not needed as it happens implicitly
the first time we call localtime() or mktime(), but in some cases
(sandboxing, chroot) this may be too late.
PR: 273807
MFC after: 3 days
Reviewed by: jrm
Differential Revision: https://reviews.freebsd.org/D41880
Removing artificial completion generator as there had been no indication
of the code being required for E810 cards. Further more it was found
that the code may have unpleasant side effects on user experience when
using ucmatose tool.
Signed-off-by: Bartosz Sobczak <bartosz.sobczak@intel.com>
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: erj@
MFC after: 1 day
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D41593
This commit fixes a couple of security vulnerabilities in the PAX writer:
1. Heap overflow in url_encode() in archive_write_set_format_pax.c
2. NULL dereference in archive_write_pax_header_xattrs()
3. Another NULL dereference in archive_write_pax_header_xattrs()
4. NULL dereference in archive_write_pax_header_xattr()
Security: No known reference yet
Obtained from: https://github.com/libarchive/libarchive/commit/1b4e0d0f9
MFC after: 3 days
[builtins][AArch64] Implement _sync out-of-line atomics
Whilst Clang does not use these, recent GCC does, and so on systems such
as FreeBSD that wish to use compiler-rt as the system runtime library
but also wish to support building programs with GCC these interfaces are
needed.
This is a light adaptation of the code committed to GCC by Sebastian Pop
<spop@amazon.com>, relicensed with permission for use in compiler-rt.
Fixes https://github.com/llvm/llvm-project/issues/63483
Reviewed By: sebpop, MaskRay
Differential Revision: https://reviews.llvm.org/D158536
Reviewed by: dim
MFC after: 1 week
Differential Revision: https://reviews.freebsd.org/D41716
Prefer libpfctl functions over direct access to the ioctl whenever
possible. This will allow subsequent removal of DIOCGETSTATUS (in 15) as
there already is an nvlist-based alternative.
MFC after: 1 week
Sponsored by: Rubicon Communications, LLC ("Netgate")
Differential Revision: https://reviews.freebsd.org/D41647
This adds the --crlf option to netcat, which triggers translation of \n
characters into \r\n sequences in the input -> network direction.
The Linux version of nc also supports this functionality with --crlf and
-C. The OpenBSD version uses -C to specify client certificates. Our
version is too old and doesn't have it, but I avoided adding -C anyway
to ease future syncs with upstream.
Attempts to upstream the feature were unsuccessful:
https://marc.info/?t=169282068500001
Differential Revision: https://reviews.freebsd.org/D41489
These tests weren't run on x86 until CI grabbed them. It turns out,
there's a sign extension bug that surfaces on x86 with char being a
signed type.
NetBSD unearthed this when they took and improved the patch, so just
grab their solution until we get to merging in the latest version of
the test.
Reported by: CI (via ngie)
Fixes: 2f489a509e ("libc: fix some overflow scenarios in vis(3)")
It was originally in contrib, and moved to usr.sbin in 6692aa840c1f; I always thought lib would make more sense but never got around to moving it.
Reviewed by: cy
Differential Revision: https://reviews.freebsd.org/D41504
Update Intel irdma driver to version 1.2.17-k
Notable changes:
- pf-reset handling improvements, including fixes in communication with if_ice(4)
- avoid racing when handling various events
- adding sw stats sysctls
- hand over pe_criterr handling from ice(4) to irdma(4)
- debug prints adjustments
- fix crash after changes in irdma_add_mqh_ifa_cb
Signed-off-by: Bartosz Sobczak <bartosz.sobczak@intel.com>
Signed-off-by: Eric Joyner <erj@FreeBSD.org>
Reviewed by: erj@
MFC after: 1 month
Sponsored by: Intel Corporation
Differential Revision: https://reviews.freebsd.org/D41425
Highlights:
- Some style fixes
- Bumped mbbuf in istrsenvisx() to MB_LEN_MAX to avoid VLAs
- mbslength cannot go negative, so make it unsigned
- Further bounds checking & fix an additional overrun, with dlen == 0
- Avoid duplicate call to wcslen(start)
GoogleTest 1.14.0 now requires C++14 to build. Change
`googletest.test.inc.mk` to reflect this requirement.
Adjust the build integration logic to handle the new version of
GoogleTest (add/remove headers/sources as needed).
Tighten down warnings via `CXXFLAGS.clang` instead of ignoring all
warnings. Some new warnings snuck in after I did my last round of fix
submissions upstream.
Also address some overlinking added in the previous version import by
removing superfluous libraries.
===============================
Expect WhenDynamicCastToTest.AmbiguousCast to fail
This change reapplies the expected failure from 1.10.0.
Ref: https://github.com/google/googletest/issues/2172
MFC after: 2 weeks
Reviewed by: asomers, emaste
Differential Revision: https://reviews.freebsd.org/D41399
Merge commit '8ef491440fcaec96f899d73e08873426c78583a4' into googletest-v1.14.0-import
The previous incarnation of this would call wcrtomb() on the destination
buffer, and only check for overflow *after* it's happened.
Additionally, the conversion error / VIS_NOLOCALE path also didn't check
for overflow, and the overflow check at the end didn't account for the
fact that we still need to write a NUL terminator afterward.
Start by only doing the multibyte conversion into mbdst directly if we
have enough buffer space to guarantee it'll fit. An additional
MB_CUR_MAX buffer has been stashed on the stack to write into if we're
cutting it close at the end of the buffer, since we don't really have a
good way to determine the length of the wchar_t without just doing the
conversion. We'll do the conversion into the buffer that's guaranteed
to fit, then copy it over if the copy won't overflow.
The byte-for-byte overflow is a little bit easier, as we simply check
for overflow with each byte written and make sure we can still NUL
terminate after.
Tests added to exercise these edge cases.
Reviewed by: des
Sponsored by: Klara, Inc.
Differential Revision: https://reviews.freebsd.org/D41328
The function ibv_query_device_ex is static inline, it is not exported
from the dso. With lld 16, which is much more picky about versioning and
undefined symbols, this becomes an error.
The ibv_register_driver driver symbol is explicitly versioned in
sources, it is non-existent in un-versioned object files.
Sponsored by: NVidia networking
MFC after: 1 week
The FreeBSD driver support for clang tested explicitly for 32-bit
Intel, MIPS, or PowerPC targets where /usr/lib32/libcrt1.o was
present to decide whether -m32 should use /usr/lib32. At jrtc27's
suggestion, simply test for a 32-bit platform rather than adding
arm to the list. Upstreamed as
3450272fc2
Bump the freebsd version to force a bootstrap build. This is one
step in adding support for -m32 on arm64.
Reviewed by: jrtc27, brooks, dim
Differential Revision: https://reviews.freebsd.org/D40943
Libarchive 3.7.0
Important changes (relevant to FreeBSD):
#1814 Do not account for NULL terminator when comparing with "TRAILER!!!"
#1818 Add ability to produce multi-frame zstd archives
#1840 year 2038 fix for pax archives on platforms with 64-bit time_t
#1860 Make single bit bitfields unsigned to avoid clang 16 warning
#1869 Fix FreeBSD builds with WARNS=6
#1873 bsdunzip ported to libarchive from FreeBSD
#1894 read support for zstd compression in 7zip archives
#1918 ARM64 filter support in 7zip archives
MFC after: 2 weeks
PR: 272567 (exp-run)
[builtins][Mips] Un-break FreeBSD build of __clear_cache
Commit 674a17e9bbe8 ("MIPS/compiler_rt: use synci to flush icache on
r6") completely removed the OS-specific guards under the guise of "For
pre-r6, we can use cacheflush libc function, which is same on Linux and
FreeBSD." However, the code in question had guards for Linux and
OpenBSD, not Linux and FreeBSD, and FreeBSD does not have a cacheflush
libc function as claimed, so this was neither the statement they
intended to make nor was it sufficient justification for making the code
completely unconditional. Whilst the upcoming FreeBSD 14 release has
dropped support for MIPS, FreeBSD 13 has support for it.
Fix this by only calling cacheflush on the OSes where it was previously
called, and not on other OSes where it either definitely isn't available
(FreeBSD) or is unknown (any other OS than the three mentioned in this
commit).
This is only needed for MFC'ing, as mips has been removed from
14-CURRENT.
PR: 271047
MFC after: immediately
Intentionally or not, but the libwrap was written in such manner that
if your /etc/hosts.allow doesn't have any domain names, neither smart
keywords like LOCAL or KNOWN, then it will not try to resolve the
client address during the hosts check. This was achieved with the
NOT_INADDR() check that matched IPv4 addresses/prefixes. Extend this
to also skip resolve if client list token looks like IPv6.
Reviewed by: philip, emaste
PR: 269456
Differential revision: https://reviews.freebsd.org/D40070
Important changes (relevant to FreeBSD):
#1840 year 2038 fix for pax archives on platforms with 64-bit time_t
#1873 bsdunzip ported to libarchive from FreeBSD
#1894 read support for zstd compression in 7zip archives
#1918 ARM64 filter support in 7zip archives
Obtained from: libarchive
Libarchive commit: ee45796171324519f0c0bfd012018dd099296336
Libarchive tag: v3.7.0
[PowerPC] Replace PPCISD::VABSD cases with generic ISD::ABDU(X,Y) node
A move towards using the generic ISD::ABDU nodes on more backends
Also support ISD::ABDS for v4i32 types using the existing signbit flip trick
PowerPC has a select(icmp_ugt(x,y),sub(x,y),sub(y,x)) -> abdu(x,y) combine that I intend to move to DAGCombiner in a future patch.
The ABS(SUB(X,Y)) -> PPCISD::VABSD(X,Y,1) v4i32 combine wasn't legal (https://alive2.llvm.org/ce/z/jc2hLU) - so I've removed it, having already added the legal sub nsw tests equivalent.
Differential Revision: https://reviews.llvm.org/D142313
This fixes a "Wasn't expecting to be able to lower this!" fatal error
when compiling graphics/opencv for PowerPC.
Requested by: pkubaj
PR: 271047
MFC after: 1 month
[clang] Reset FP options before template instantiation
AST nodes that may depend on FP options keep them as a difference
relative to the options outside the AST node. At the moment of
instantiation the FP options may be different from the default values,
defined by command-line option. In such case FP attributes would have
unexpected values. For example, the code:
template <class C> void func_01(int last, C) {
func_01(last, int());
}
void func_02() { func_01(0, 1); }
#pragma STDC FENV_ACCESS ON
caused compiler crash, because template instantiation takes place at the
end of translation unit, where pragma STDC FENV_ACCESS is in effect. As
a result, code in the template instantiation would use constrained
intrinsics while the function does not have StrictFP attribute.
To solve this problem, FP attributes in Sema must be set to default
values, defined by command line options.
This change resolves https://github.com/llvm/llvm-project/issues/63542.
Differential Revision: https://reviews.llvm.org/D154359
Requested by: pkubaj
PR: 265755, 265758
MFC after: 1 month
See commit 8fad2cda93 ("bsd.compat.mk: Provide new CPP and sub-make
variables") for the context behind this change.
Reviewed by: brooks, jhb
Differential Revision: https://reviews.freebsd.org/D40923
[Clang] Show type in enum out of range diagnostic
When the diagnostic for an out of range enum value is printed, it
currently does not show the actual enum type in question, for example:
v8/src/base/bit-field.h:43:29: error: integer value 7 is outside the valid range of values [0, 3] for this enumeration type [-Wenum-constexpr-conversion]
static constexpr T kMax = static_cast<T>(kNumValues - 1);
^
This can make it cumbersome to find the cause for the problem. Add the
enum type to the diagnostic message, to make it easier.
Reviewed By: aaron.ballman
Differential Revision: https://reviews.llvm.org/D152788
PR: 271047
MFC after: 1 month