Age | Commit message (Collapse) | Author |
|
|
|
|
|
ok kettenis@
|
|
|
|
ok hackroom
tested by plenty
|
|
ok hackroom
tested by plenty
|
|
binary to /usr/bin and do the same with the manpage as well and make sure
that we only build llvm-ar on architectures where it is actually enabled
discussed with deraadt@
|
|
and libLLVMLibDriver;
switch LLD_ARCHs to llvm-ar(1) by skipping the installation of binutils' ar(1)
and linking llvm-ar(1) to ar(1)
tested on amd64, i386, arm64 and mips64
ok patrick@, kettenis@
|
|
readelf to display them in a way compatible with llvm-readelf, including
the --raw-relr debugging option
ok kettenis@
|
|
Our emulated TLS implementation relies on local state (e.g. for the pthread
key), and if we duplicate this state across different shared libraries,
accesses to the same TLS variable from different shared libraries will yield
different results (see https://github.com/android/ndk/issues/1551 for an
example). __emutls_get_address is the only external entry point for emulated
TLS, and by making it default visibility and weak, we can rely on the dynamic
linker to coalesce multiple copies at runtime and ensure a single unique copy
of TLS state. This is a best effort;
Also bump the libc++abi minor because now it picks up the __emutls_get_address
symbol.
ok kettenis@
|
|
originally from Tony Cook in
skip using gcc brace groups for STMT_START/END
7169efc77525df70484a824bff4ceebd1fafc760
looks fine millert@ ok afresh1@
|
|
looking good sthen@, Great! bluhm@
|
|
The IRIX-specific extension of the PT_DYNAMIC segment is not needed
by the dynamic linker on OpenBSD/mips64. Disable it so that the .dynamic
section stays at the start of the PT_DYNAMIC segment even when .dynstr,
.dynsym or .hash precedes .dynamic in the ELF file. This enables
Binutils 2.17 tools, such as strip(1), rewrite executables and shared
libraries that have been produced by LLD.
OK kettenis@
|
|
"go for it" kettenis@
|
|
Fixes at least ports/devel/libffi as used by ports/lang/python/2.
ok kettenis@
|
|
Not linking it to the build yet.
"go for it" mortimer@
|
|
functions
ok deraadt@
different versions tested by jca@ naddy@ sthen@
|
|
With this change, passing -fprofile-instr-generate -fcoverage-mapping
when building programs will attempt linking against libclang_rt.profile.a.
Please note that we do not ship the library yet.
OK jca@
|
|
Recommended by AMD white paper Software Techniques for Managing
Speculation on AMD Processors (9.17.20) mitigation V2-1.
Pointed out by bluhm@. ok bluhm@ kettenis@
|
|
|
|
Clang uses runtime libraries for some advanced features like
sanitizers. Different systems have different preferences about file
placement. OpenBSD with this change would use this name for ASan:
/usr/lib/clang/11.1.0/lib/libclang_rt.asan.a
So far, no libraries are shipped, only their eventual location is chosen.
Discussed with deraadt@ and kettenis@, OK mortimer@
|
|
Suggested by jca@ that kettenis@ looks at it first.
|
|
Tested by doing a full system build locally. Will work with brad@ for
upstream-ing.
OK patrick & mortimer
|
|
Add powerpc to LLD_ARCH. Remove -Wl,-relax flag; ld.bfd needed it,
but ld.lld can't take this flag.
To avoid linker errors on macppc, you need a snapshot where
/usr/bin/ld is LLD.
ok kettenis@ deraadt@
|
|
This is upstream commit
https://github.com/Perl/perl5/commit/c1a937fef07c061600a0078f4cb53fe9c2136bb9
Quoting upstream:
Without this fix, Encode::ConfigLocal can be loaded from a path
relative to the current directory, because the || operator will
evaluate @inc in scalar context, putting an integer as the only
value in @inc.
Addresses CVE-2021-36770
|
|
tools, and it also provides command line options and thus pollutes the name
space. This fixes duplicate command line options in llvm-objdump.
Discussed with jsg@
|
|
ok kettenis@ sthen@
|
|
GCC versions will sometimes emit it, and we have stumbled upon it a few times
in ports.
Upstream commit: https://reviews.llvm.org/D95994
GCC bug report: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=98946
ok kettenis@
|
|
Instead of "<unknown>: f3". ok kettenis@
|
|
Instead of swapping registers around, we can just use the REV version of
the same instruction, which has the same effect but encodes differently and
does not result in return bytes in the binary. This reduces the number
of xchg instructions resulting from gadget fixing.
Prompted by ratchov@, with input from millert@ and sthen@.
ok sthen@
|
|
base-gcc always errored out when -Werror was passed and -Wuninitialized
triggered, even when -Wno-error=uninitialized was passed.
Deemed correct by Miod
|
|
|
|
defines because we need it now
from https://reviews.llvm.org/D91784
ok mlarkin kettenis
|
|
ok hackroom
tested by plenty
|
|
ok deraadt@ drahn@
|
|
ok jsg drahn
|
|
and Readline.
feedback and okay afresh1@
|
|
discussion with espie kettenis jsg
|
|
|
|
ok mortimer@
|
|
ok mortimer@
|
|
ok mortimer@
|
|
Pointed out by deraadt@, this is not what that that incomplete table is for.
While the code has been there for several years, I haven't found a use for it,
which is good because it would have been wrong.
While here, update my name.
|
|
With some build plumbing help from jsg@
ok kettenis@ sthen@
|
|
Based on NetBSD implementation, without the libelf dependency. Architectures
which have libunwind use libunwind, and others use a stub implementation
that does nothing since __builtin methods are unreliable.
Much feedback and help from jca@.
ok kettenis@ sthen@
|
|
This 'requirement' was due to a bug in dtors which has been corrected.
reminded by jsg@
|
|
libc++abi: Pure virtual function called!
By statically linking clang (and *-tblgen) this problem is worked around
at least for now.
|
|
Uncertain if this is 100% correct, however this enables self hosting
using clang.
|
|
ok hackroom
tested by plenty
|
|
default to -no-relax.
ok drahn@
|