Apart from bumping the toolchain Clang's and port's version, this commit
completely overhauls the way LLVM toolchain is built.
First, it gets rid of a complicated two-stage process of first compiling
clang and compiler-rt builtins and then building libunwind, libc++abi,
and libc++ -- it is possible to create a complete cross-compilation
toolchain in a single CMake invocation with a modern LLVM. Moreover, the
old method was inherently unsupported and subtly broken.
Next, it utilizes full potential of the Stubs "framework". Now we are
even able to compile Clang with -Wl,-z,defs which makes one of the
patches obsolete and the whole installation less error-prone. Note that
it comes at a cost of complicating the bootstrap process on a completely
novel architecture but this hopefully won't happen often.
Lastly, it fixes handling of the -no*lib* family of flags in the
Serenity LLVM driver and correctly uses -nostartfiles in conjunction
with stubs to make necessary CMake configure-time checks succeed.
This commit adds minimal support for compiler-instrumentation based
memory access sanitization.
Currently we only support detection of kmalloc redzone accesses, and
kmalloc use-after-free accesses.
Support for inline checks (for improved performance), and for stack
use-after-return and use-after-return detection is left for future PRs.
This is a minor bugfix release, which to my knowledge contains nothing
of importance to us. However, there is one QoL change to our patches.
We no longer force `-fpic` in the compiler driver, and instead use the
`--enable-default-pie` configure option to generate position-independent
code suitable for executables. For building shared libraries, the
`-fpic` flag must be specified explicitly.
This version contains my patch that adds support for the proprietary
VideoCore mailbox message for reading the kernel command line, so
patches aren't needed anymore.
The BFD linker requires the `COMMONPAGESIZE` emulation parameter to be
set in order to enable RELRO support for AArch64. As we are adding a
custom `emulparams` file anyways, let's also tell LD that our ELF
interpreter is called `/usr/lib/Loader.so`.
This commit also removes some vestigial references to i686 SerenityOS.
The one in `gas/configure.tgt` is still needed, as it also handles
x86_64.
This is a mostly straight-forward rebase of our patches on top of
13.1.0. The spec files needed a change, as GCC no longer supports STABS
debug information, but we were building GCC with support for it.
Highlights of this release include static `operator()`, The Equality
Operator You Are Looking For and extended `constexpr` support.
aarch64 required this to successfully build, but the actual fix to that
is to just make sure that crtbeginS and crtendS get built as a part of
the toolchain.
This partially reverts commit c18c84dbfd.
Without this tag, executables that link libgcc_s.so.1, which is every
executable, end up with the host path to libgcc_s.so.1 as the path for
the dynamic dependency (DT_NEEDED) of the executable. By making sure
that the libgcc_s.so.1 shared object includes the DT_SONAME tag, the
path to libgcc_s.so.1 dynamic dependency will only contain the basename,
instead of the absolute host path.
QEMU 7.1 was released on August 30th.
Release Notes: https://wiki.qemu.org/ChangeLog/7.1
The patch 'Toolchain/Patches/qemu-cf-protection-none.patch' (or similar)
has been upstreamed so it can be safely removed.
I originally missed this while handling the upgrade to LLVM 15 (as it
only affects the on-serenity port), so the patch gets to be here with a
bit of a delay.
Co-Authored-By: sin-ack <sin-ack@users.noreply.github.com>
There was a typo in one of the spec files which resulted in us not
building softfp support for libgcc. Additionally we were missing flags
to build libgcc_s. This patch also makes sure we're not trying to
link against crtbeginS.o and crtendS.o. This is part of a larger effort
to at least get the userland to build at all.
Now that the lower pages can be unmapped and more of the virtual
address range is available to us, we can actually use the default
mapping address of x86_64 again.
This reverts commit 292398b585.
This shouldn't cause any breaking changes, so a toolchain rebuild is not
required.
As per Hendiadyoin's request, math errno is disabled by default, which
should enable some extra compiler optimizations in LibGL and LibSoftGPU
code that uses math functions heavily.
Co-Authored-By: Ali Mohammad Pur <mpfard@serenityos.org>
This release brings support for various C++23 constructs like `if
consteval` and multidimensional subscript operators. Vectorization is
now enabled for O2 too, and `-ftrivial-auto-var-init` has been added
which can help us find and prevent security issues coming from
uninitialized variables.
Toolchain/Patches/gcc.patch is now significanly smaller as some unused,
autoconf-generated code has been removed.
Add a patch to let llvm's InstrProfiling modules know serenity supports
all the Unix-y features required to make -fprofile-instr-generate and
-fcoverage-mapping work properly on target.
Besides a version bump, the following changes have been made to our
toolchain infrastructure:
- LLVM/Clang is now built with -march=native if the host compiler
supports it. An exception to this is CI, as the toolchain cache is
shared among many different machines there.
- The LLVM tarball is not re-extracted if the hash of the applied
patches doesn't differ.
- The patches have been split up into atomic chunks.
- Port-specific patches have been integrated into the main patches,
which will aid in the work towards self-hosting.
- <sysroot>/usr/local/lib is now appended to the linker's search path by
default.
- --pack-dyn-relocs=relr is appended to the linker command line by
default, meaning ports take advantage of RELR relocations without any
patches or additional compiler flags.
The formatting of LLVM port's package.sh has been bothering me, so I
also indented the arguments to the CMake invocation.
I noticed after upgrading my machine that the QEMU is no longer building
due to GCC enabling `-fcf-protection` by default, even for targets that
don't support it.
The included patch came from the QEMU development list, but hasn't be
included in any patch releases at the time of writing.
https://lore.kernel.org/all/20220208211937.79580-1-vineetg@rivosinc.com/
Until QEMU patches, lets fix it on our end by patching before we build.
Although we handle it in the linker spec file, gcc will actually reject
this argument unless it's also enabled in the option file.
CMake adds this flag if the minimum required version is 3.3 or less (see
CMP0065), so old projects would fail to compile because of this
unrecognized option.
This commit backports the LLVM commit that adds support for the
`--update-section` flag to llvm-objcopy. We use this feature of GNU
objcopy to embed the symbol map in the kernel.
The corresponding LLVM Phabricator Differential Revision can be found
here: https://reviews.llvm.org/D112116
This patch is identical to the upstream commit, except for two hunks
that had to be changed as they didn't apply cleanly.
Discord user aesophor pointed out that the GCC toolchain fails to build
on macOS, and traced the issue back to 41ea37f2, which is the latest
change to `gcc.patch`. Similarly, when I tried to run BuildIt.sh in the
`--dev` mode, `git apply` complained about the patch being malformed.
I regenerated the patch by manually applying the changes of 41ea37f2 on
top of a known good GCC source tree, and I sent the new file to them.
They reported that this fixed the build issue they were having.
