RISC-V has no support for subword atomic operations; code currently
generates libatomic library calls.
This patch changes the default behavior to fast inline subword atomic
calls that do not require libatomic.
Signed-off-by: Tianling Shen <cnsztl@immortalwrt.org>
The glibc build was failing with this error message:
aclocal.m4:6: error: Exactly version 2.69 of Autoconf is required but you have 2.71
aclocal.m4:6: the top level
The autoconf run on glibc is not needed, just remove it.
Fixes: #12631
Fixes: 030447b8f4 ("tools/autoconf: bump to 2.71")
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Release notes:
new features:
- large dns record lookups via tcp fallback
- new getaddrinfo EAI_NODATA result to distinguish NODATA/NxDomain
- support for new RELR compressed format for relative relocations
- sysconf keys for querying signal stack size requirements
- real vfork on riscv64
performance:
- mallocng no longer uses MADV_FREE (high performance cost, little gain)
- vdso clock_gettime is supported once again on 32-bit arm
compatibility:
- gethostbyname family now distinguishes NO_DATA from HOST_NOT_FOUND
- res_send now works with caller-provided edns0 queries
- arpa/nameser.h RR types list is now up-to-date
- previously-missing POSIX confstr keys have been added
- mntent interfaces now accept missing fields
- alt signal stack, if any, is now used for internal signals
- the LFS64 macros are no longer exposed without _LARGEFILE64_SOURCE
- memmem (POSIX-future) is now exposed in default feature profile
- pthread_atfork now admits calls from an application-provided malloc
- debugger tracking of shared libraries now works on MIPS PIE binaries
- sendmsg now supports up to SCM_MAX_FD fds in SCM_RIGHTS messages
bugs fixed:
- gethostbyname[2]_r wrongly returned nonzero (error) on negative result
- parallel v4/v6 address queries could fail on query id collisions
- spurious getaddrinfo/AI_ADDRCONFIG failures due to errno clobbering
- dns search domains ending in dot (including lone dot) broke lookups
- ipv6 servers in resolv.conf broke lookups on systems with v6 disabled
- systems with bindv6only failed to query both v4 and v6 nameservers
- res_mkquery mishandled consecutive final dots in name
- res_send could malfunction for very small answer buffer sizes
- resolver dns backend accepted answers with wrong (A vs AAAA) RR type
- getservbyport_r returned junk or ENOENT (vs ERANGE) on buffer size errors
- dns result parsing of malformed responses could process uninitialized data
- freopen didn't reset stream orientation (byte/wide) & encoding rule
- fwprintf didn't print most fields on open_wmemstream FILEs
- wide printf %lc ignored field width
- wide printf erroneously processed %n after encoding errors
- use of wide printf %9$ argument slot overflowed undersized buffer
- swprintf malfunctioned on nul character in output
- strverscmp ordered digit sequences vs nondigits incorrectly
- timer_create/SIGEV_THREAD failure leaked the thread
- semaphores were subject to missed-wake under certain usage patterns
- several possible rare deadlocks with lock handling at thread exit
- several possible rare deadlocks with aio and multithreaded fork
- dynamic linker relro processing was broken on archs w/variable pagesize
- async cancellation could run cancellation handlers in invalid context
- pthread_detach was wrongly a cancellation point in rare race code path
- use-after-close/double-close errors in mq_notify error paths
- mq_notify event thread wrongly ran with signals unmasked
- wcs{,n}cmp, wmemcmp returned wrong results when difference overflowed
- accept4, pipe2, and dup3 handled unknown flags wrong in fallback cases
- CPU_SETSIZE macro had wrong unit
- select fallback for pre-time64 kernels truncated timeout (vs clamping)
arch-specific bugs fixed:
- x32 new socketcalls took fallback path due to pointer sign extension
- x32 wait4 didn't fill rusage structure (time64 regression)
- x32 semtimedop mismatched timespec ABI with kernel (time64 regression)
- sigaction signal mask was bogus on or1k, microblaze, mips, and riscv
- powerpc-sf longjmp asm clobbered value argument
- or1k poll function passed timeout to syscall in wrong form
Removed upstreamed:
- 800-mips_pie_debug.patch
Manually rebased:
- 600-nftw-support-common-gnu-extension.patch
Signed-off-by: Linhui Liu <liulinhui36@gmail.com>
some targets do not define the _REDIR_TIME64 macro resulting in a
build error regression since ddfe5678a4
fix by checking if the macro is defined
[1] https://github.com/openwrt/openwrt/issues/12587Fixes: #12587
Fixes: ddfe5678a4 ("fortify-headers: fix inconsistent time_t version of ppoll")
Signed-off-by: Georgi Valkov <gvalkov@gmail.com>
The purpose of this is to remove all of the compatibility code that is
normally compiled into glibc for kernels < 5.15 from our build. Only
Linux kernels >= 5.15 will be supported.
Signed-off-by: Nick Hainke <vincent@systemli.org>
While improving access path analysis a typo happened. Now it can happen
that gcc misscompiles. The patch is fixing the issue. However, also
other gcc versions 10.2+ are affected. They also should be bumped or the
fix should be backported.
For more bug information have a look at:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=109585
Signed-off-by: Nick Hainke <vincent@systemli.org>
LTO object files will now be compressed using zstd.
Compressing debug sections with zstd will have to wait for GCC 13,
which adds support for -gz=zstd.
NOTE: wiping the ccache is strongly recommended, not doing so might
yield build error later on:
"lto1: internal compiler error: original not compressed with zstd"
Signed-off-by: Andre Heider <a.heider@gmail.com>
This was build tested with all core packages on all targets
successfully.
This was run tested on the following systems:
* lantiq/xrx200 musl
* x86/64 musl
* x86/64 glibc
* malta/be
The size of some of the initramfs images from lantiq/xrx increased by
240 bytes.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Add the patches with real changes from the binutils 2.40 stable branch.
I am not aware that we ran into any of these problems, but I think it is
better to take the existing stable patches.
They were exported like this:
git format-patch binutils-2_40...origin/binutils-2_40-branch
I removed the patches changing the version numbers and updating the
translations only.
I removed the following patches:
*Automatic-date-updat*
001-Re-enable-development.-Update-version-to-2.40.0.patch
004-Updated-translations-for-the-gas-and-binutils-sub-di.patch
015-Updated-Swedish-translation-for-the-binutils-sub-dir.patch
027-Updated-Swedish-translation-for-the-binutils-sub-dir.patch
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The utils subtarget was used for uClibc and eglibc
which are both obsolete and replaced in Openwrt by mainline
musl and glibc, which do not have a utils Makefile target.
Ref: 5d19a38d8 ("toolchain: sync eglibc headers/build split with uclibc changes")
Fixes: 63fb17520 ("toolchain: remove uClibc-ng")
Suggested-by: Andre Heider <a.heider@gmail.com>
Signed-off-by: Michael Pratt <mcpratt@pm.me>
Reorganize dependencies between toolchain targets to save a line.
This also puts them in a linear order which is easier to read.
Signed-off-by: Michael Pratt <mcpratt@pm.me>
Use the gcc/final/install target stampfile as the prerequisite
for the toolchain/compile target stampfile.
The previous filename was also incorrect,
the stampfile name format was changed (see reference)
but was never updated here.
The issue was never noticed
because the usage of timestamp.pl was broken
(fixed in previous commit)
causing the extra prereq files passed to 'call stampfile'
to never get checked at all.
Fixes: 5149ed151 ("gcc: split up the build process into three distinct stages")
Signed-off-by: Michael Pratt <mcpratt@pm.me>
This was apparently introduced to recreate the toolchain (wipe
staging_dir/toolchain*, but keep build_dir/toolchain*, followed by a
`make toolchain/compile`).
But it leaves leftovers and causes re-links to happen at src_install phase,
because of the changed paths, possibly adding yet another source of issues.
With the prior commits removing various hacks related to the "initial"
folder we can remove installing it twice altogether.
The recreated toolchain is exactly the same as before.
Signed-off-by: Andre Heider <a.heider@gmail.com>
Add it to the existing list to fix it up.
Found by mold:
toolchain-x86_64_gcc-12.2.0_glibc/lib/libm.so:4: GROUP ( /lib/libm.so.6 AS_NEEDED ( /lib/libmvec.so.1 ) )
^ library not found: /lib/libm.so.6
Signed-off-by: Andre Heider <a.heider@gmail.com>
A modern glibc only links against libgcc_eh for its tests or when building it
static, which doesn't happen here.
Reverts a hunk of: a3edea1b "add support for alternative C libraries..."
Signed-off-by: Andre Heider <a.heider@gmail.com>
This reverts commit 9d1b619cb1.
No Linaro binutils are supported anymore, the leftover removal commit
missed one line.
Fixes: b648e1c0 "binutils: remove 2.25.1 support and leftovers of older versions"
Signed-off-by: Andre Heider <a.heider@gmail.com>
In [0] support for ARC specific binutils were added, which was later removed
again [Fixes:].
Upstream binutils never had an extlib.
[0] 8b5d644b "toolchain: add support of ARC architecture"
Fixes: 8568dcd9 "toolchain: Remove binutils 2.29 for ARC targets"
Signed-off-by: Andre Heider <a.heider@gmail.com>
This updates to glibc to version 2.37.
Signed-off-by: Linhui Liu <liulinhui36@gmail.com>
[Update to current git commit]
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Instead of having two different ways to pass flags to the gcc build
process, add them as configure args, which is a reliable way to let
gcc pass them around to its various pieces.
Also add CXXFLAGS, since gcc started to use c++ for itself recently
(~10 years ago now).
Signed-off-by: Andre Heider <a.heider@gmail.com>
Spell out what we want to enable or disable. This prevents host libs to leak in,
so everyone get the same feature set.
Signed-off-by: Andre Heider <a.heider@gmail.com>
Spell out what we want to enable or disable. This prevents host libs to leak in,
so everyone get the same feature set.
Signed-off-by: Andre Heider <a.heider@gmail.com>
Kconfig docs say:
> The default value deliberately defaults to 'n' in order to avoid
> bloating the build.
Apply this rule everywhere, to avoid more cloning of bad examples
Signed-off-by: Tony Butler <spudz76@gmail.com>
Fix DT_DEBUG handling on MIPS in musl libc.
With this change gdb will load the symbol files for shared libraries on MIPS too.
This patch was taken from this thread: https://www.openwall.com/lists/musl/2022/01/09/4
Signed-off-by: Hauke Mehrtens <hmehrtens@maxlinear.com>
Use STAGING_DIR_HOST to reference staging host directory instead of
hardcoding it to default path.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Use STAGING_DIR_HOST to reference staging host directory instead of
hardcoding it to default path.
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
Config evaluation require default with if to be put before the generic
default config with no condition. Putting the default config before any
conditional default results in always selecting the non conditional one.
This results in the version be hardcoded to gcc 12 even if gcc 11 is
selected in the Advanced build options.
Fix this by putting the gcc 12 default option as last after ANY
conditional default config.
Fixes: d9de5252a4 ("toolchain/gcc: switch to version 12 by default")
Signed-off-by: Christian Marangi <ansuelsmth@gmail.com>
This was build tested with all core packages on all targets
successfully.
Most packages from the feed are also building fine.
This was run tested on the following systems:
* lantiq/xrx200 musl
* pistachio/generic musl
* sunxi/cortexa53 musl
* x86/64 musl
* x86/64 glibc
* armvirt/64 musl
The size of the images stays more or less the same for MIPS BE and
aarch64. I haven't tested other architectures.
With GCC 11 I got these sizes for lantiq/xrx200:
7,219,848 openwrt-lantiq-xrx200-tplink_tdw8970-initramfs-kernel.bin
7,472,208 openwrt-lantiq-xrx200-tplink_tdw8970-squashfs-sysupgrade.bin
With GCC 12 I got these sizes for lantiq/xrx200:
7,217,355 openwrt-lantiq-xrx200-tplink_tdw8970-initramfs-kernel.bin
7,406,674 openwrt-lantiq-xrx200-tplink_tdw8970-squashfs-sysupgrade.bin
The sysupgrade image is probably padded. The initramfs image is 0.03% smaller.
With GCC 11 I got these sizes for armvirt/64:
4,143,943 openwrt-armvirt-64-default-rootfs.tar.gz
10,887,176 openwrt-armvirt-64-Image
24,911,880 openwrt-armvirt-64-Image-initramfs
4,141,572 openwrt-armvirt-64-rootfs.cpio.gz
4,255,854 openwrt-armvirt-64-rootfs-ext4.img.gz
3,391,178 openwrt-armvirt-64-rootfs-squashfs.img.gz
With GCC 12 I got these sizes for armvirt/64:
4,142,778 openwrt-armvirt-64-default-rootfs.tar.gz
10,887,176 openwrt-armvirt-64-Image
24,911,880 openwrt-armvirt-64-Image-initramfs
4,138,105 openwrt-armvirt-64-rootfs.cpio.gz
4,255,463 openwrt-armvirt-64-rootfs-ext4.img.gz
3,390,390 openwrt-armvirt-64-rootfs-squashfs.img.gz
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
The version check which sets GCC_VERSION_FILE to the correct value only
worked when the advanced options menu was active and not when it was not
active.
Thank you Tony Butler for the fix.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
ChangeLog:
Version 2.16.01
_This is a documentation update release only._
(*) Fix the creation of the table of contents in the HTML version of
the documentation.
Version 2.16
(*) Support for the `rdf' format has been discontinued and all the
RDOFF utilities has been removed.
(*) The `--reproducible' option now leaves the filename field in the
COFF object format blank. This was always rather useless since
it is only 18 characters long; as such debug formats have to
carry their own filename information anyway.
(*) Fix handling of MASM-syntax reserved memory (e.g. `dw ?') when
used in structure definitions.
(*) The preprocessor now supports functions, which can be less
verbose and more convenient than the equivalent code implemented
using directives. See section 4.4.
(*) Fix the handling of `%00' in the preprocessor.
(*) Fix incorrect handling of path names affecting error messages,
dependency generation, and debug format output.
(*) Support for the RDOFF output format and the RDOFF tools have
been removed. The RDOFF tools had already been broken since at
least NASM 2.14. For flat code the ELF output format
recommended; for segmented code the `obj' (OMF) output format.
(*) New facility: preprocessor functions. Preprocessor functions,
which are expanded similarly to single-line macros, can greatly
simplify code that in the past would have required a lengthy
list of directives and intermediate macros. See section 4.4.
(*) Single-line macros can now declare parameters (using a `&&'
prefix) that creates a quoted string, but does _not_ requote an
already quoted string. See section 4.2.1.
(*) Instruction table updated per public information available as of
November 2022.
(*) All warnings in the preprocessor have now been assigned warning
classes. See appendix A.
(*) Fix the invalid use of `RELA'-type relocations instead of `REL'-
type relocations when generating DWARF debug information for the
`elf32' output format.
(*) Fix the handling `at' in `istruc' when the structure contains
local labels. See section 5.9.2.
(*) When assembling with `--reproducible', don't encode the filename
in the COFF header for the `coff', `win32' or `win64' output
formats. The COFF header only has space for an 18-character
filename, which makes this field rather useless in the first
place. Debug output data, if enabled, is not affected.
(*) Fix incorrect size calculation when using MASM syntax for non-
byte reservations (e.g. `dw ?'.)
(*) Allow forcing an instruction in 64-bit mode to have a (possibly
redundant) REX prefix, using the syntax `{rex}' as a prefix.
(*) Add a `{vex}' prefix to enforce VEX (AVX) encoding of an
instruction, either using the 2- or 3-byte VEX prefixes.
(*) The `CPU' directive has been augmented to allow control of
generation of VEX (AVX) versus EVEX (AVX-512) instruction
formats, see section 7.11.
(*) Some recent instructions that previously have been only
available using EVEX encodings are now also encodable using VEX
(AVX) encodings. For backwards compatibility these encodings are
not enabled by default, but can be generated either via an
explicit `{vex}' prefix or by specifying either `CPU LATEVEX' or
`CPU NOEVEX'; see section 7.11.
(*) Document the already existing `%unimacro' directive. See section
4.5.12.
(*) Fix a code range generation bug in the DWARF debug format
(incorrect information in the `DW_AT_high_pc' field) for the ELF
output formats. This bug happened to cancel out with a bug in
older versions of the GNU binutils linker, but breaks with other
linkers and updated or other linkers that expect the spec to be
followed.
(*) Fix segment symbols with addends, e.g. `jmp _TEXT+10h:0' in
output formats that support segment relocations, e.g. the `obj'
format.
(*) Fix various crashes and hangs on invalid input.
Signed-off-by: Linhui Liu <liulinhui36@gmail.com>
This was build tested with all core packages on all targets
successfully.
This was run tested on the following systems:
* lantiq/xrx200 musl
* sunxi/cortex53 musl
* x86/64 musl
* x86/64 glibc
Some trusted firmware arm builds needed some fixes to build with
binutils 2.39, this was merged before.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
Add the patches with real changes from the binutils 2.39 stable branch.
I am not aware that we ran into any of these problems, but I think it is
better to take the existing stable patches.
They were exported like this:
git format-patch binutils-2_39...origin/binutils-2_39-branch
I removed the patches changing the version numbers only.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
glibc 2.36 changed the definition of enum fsconfig_command, it now
collides with the same definition from sys/mount.h. Remove the include
of linux/fs.h This still compiled with musl too.
This backports a patch which is already in the stable branch of GCC 11
and GCC 12.
Signed-off-by: Hauke Mehrtens <hauke@hauke-m.de>
