Age | Commit message (Collapse) | Author |
|
It has been almost 3 months since the upgrade to gcc 10. Switch back to
relying on gcc default configuration (-fno-common).
This reverts OE-Core commit 951e859b1e8297970278c539e989b8a6d06a9cb3.
Signed-off-by: Jacob Kroon <jacob.kroon@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
For time being change back to -fcommon as default, helps us iron
out other issues, eventually this should be removed as we fix the
packages to work with -fno-common
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
OE uses libiberty from binutils, since its properly compiled as pic
archive and applications and other libraries needing libiberty can
properly link with it.
With this option applied, explicit delete of libiberty headers and
libraries is not required in install step, since they wont get installed
in first place.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
While the gnu style for symbol versioning is the most usual,
--enable-symvers[=style] can be provided several values,
gnu, gnu-versioned-namespace, darwin, darwin-export, and sun,
depending on users needs.
Introduce the SYMVERS_CONF variable to allow the user to
configure the symbol versioning in shared libraries.
Signed-off-by: Alejandro Enedino Hernandez Samaniego <alejandro@enedino.org>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Let the patch trigger based on target triplet instead of passing via
configure, this lets gcc compile for 64bit otherwise it ends up with
libgcc build errors
error: unable to emulate 'TF'
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Java support was removed in upstream gcc 7.
Signed-off-by: Adrian Bunk <bunk@stusta.de>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Separated out from the previous commit for clarity, this simply drops
all the -inital pieces of gcc which are no longer needed after the
previous commit.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
We need a libgcc to build glibc. Tranditionally we therefore build
a non-threaded and non-shared compiler (gcc-cross-initial), then use
that to build libgcc-initial which is used to build glibc which we can
then build gcc-cross and libgcc against.
Firstly, we can drop the glibc dependency from gcc-cross, *if* we make
two changes:
a) specify the minimum glibc version to support in a configure option
b) create a dummy limits.h file so that later when glibc creates one,
the headers structure has support for it. We can do this with a simple
empty file
Once gcc-cross is libc independent, we can use it to build both
libgcc-initial and then later libgcc.
libgcc-initial is tricky as we need to imitate the non-threaded and
non-shared case. We can do that by hacking the threading mode back to
"single" even if gcc reports "posix" and disable libc presence for the
libgcc-intial build. We have to create the dummy limits.h to avoid
compiler errors from a missing header.
glibc will fail to link with libgcc-initial due to a missing "exception
handler" capable libgcc (libgcc_eh.a). Since we know glibc doesn't need
any exception handler, we can safely symlink to libgcc.a.
With those changes, gcc-cross can be used in all places and we only need
one build of gcc for each architecture.
For some reason ifunc was being disabled on mips prior to these changes
but afterwards became enabled but caused assertion failures. This is
therefore disabled until we can debug that.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
GCCPIE flag which is empty by default adds "--enable-default-pie"
configure option for harderned distros
We do not require to add -fpie -pie flag externally anymore
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
uclibc support was removed a while ago and musl works much better. Start to
remove the various overrides and patches related to uclibc which are no longer
needed.
uclibc support in a layer would still be possible. I have strong reasons to
believe nobody is still using uclibc since patches are missing and I doubt
the metadata even parses anymore.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
We only build one gcc-cross per architecture and having target specific
flags means gcc-cross would rebuild. These flags are really for the
on target case, so they should be set in gcc-target only.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
getVar() now defaults to expanding by default, thus remove the True
option from getVar() calls with a regex search and replace.
Search made with the following regex: getVar ?\(( ?[^,()]*), True\)
Signed-off-by: Joshua Lock <joshua.g.lock@intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
Add support for MIPS Release 6 ISA
Signed-off-by: Zubair Lutfullah Kakakhel <Zubair.Kakakhel@imgtec.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This adds the correct support for initfini-array which replaces .init
and .fini with .init-array and .fini-array. There is no appreciable
size difference with this change.
The change is needed since configure will not correctly detect support
when building cross-compilers.
Signed-off-by: Haitao Huang <haitao.huang@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
Configuring gcc with --enable-target-optspace (which causes gcc to
append "-g -Os" to the default CFLAGS_FOR_TARGET and so force libgcc
etc target libraries to always be optimised for size) dates back to
the very first commit in oe-core git in 2005 (for gcc 3.4.3).
Configuring gcc with --enable-target-optspace is not done widely
elsewhere (it's not used for Ubuntu or Fedora host gcc, the Linaro
binary toolchain or in Buildroot since early 2015). Sometime around
gcc 4.5.x it caused problems for powerpc and so was disabled for that
architecture:
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=43810
This patch removes --enable-target-optspace completely (ie powerpc is
no longer a special case) and allows optimisation of libgcc etc to be
controlled directly by the flags present in TARGET_CFLAGS.
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The Cortex M1, M3 and R4 CPU tuning files are poorly tested (if at
all). They have no obvious users either inside or outside oe-core.
Until OE officially gains support for CPUs without an MMU, these
tuning files are probably better maintained outside of oe-core (e.g.
in a separate meta-nommu layer).
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
Signed-off-by: Andre McCurdy <armccurdy@gmail.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
The gcc default, bss-plt, will cause errors when using the prelinker. All
other distributions that I am aware of are using the the secure-plt. For an
explanation of the differences, the gcc docs:
Current PowerPC GCC accepts a `-msecure-plt' option that generates code
capable of using a newer PLT and GOT layout that has the security
advantage of no executable section ever needing to be writable and no
writable section ever being executable. PowerPC ld will generate this
layout, including stubs to access the PLT, if all input files (including
startup and static libraries) were compiled with `-msecure-plt'.
`--bss-plt' forces the old BSS PLT (and GOT layout) which can give
slightly better performance.
The security of the new PLT and ability to run the prelinker outweigh
any performance penalty.
The secure-plt is enabled by default. The old bss-plt can be enabled by
selecting 'bssplt' in the DISTRO_FEATURES.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
The gcc-intermediate had been gone, so remove EXTRA_OECONF_INTERMEDIATE.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
It is not needed any more since sed-native had been dropped.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
|
|
[YOCTO #7230]
In certain system configurations TRANSLATED_TARGET_ARCH will not
expand in the right order for gcc-cross-candian-mips64n32 to be
generated properly.
This will cause SDKs to fail to generate properly.
Changing the global definition of TRANSLATED_TARGET_ARCH always
expands the ABIEXTENSION, which causes the OVERRIDES to pick it up
as well. This effectively defines a new class of overrides for the 'n32'.
The side effect is that we need to duplicate some mips64 overrides, and
redefine others that were previously 'n32' or 'mips64' exclusive to have
the correct semantics.
Signed-off-by: Mark Hatle <mark.hatle@windriver.com>
|
|
The current implementation of shared work for gcc is at best confusing. It relies
on the fetch/unpack/patch tasks having exactly the same stamps and if this gets
broken for some reason, its hard to figure out what the problem is. It also
leads to complex code in bitbake.
The benefits of shared work for gcc are clear but a better approach is needed. This
patch adjusts things so that a single new recipe (gcc-source) provides the
fetch/unpack/patch/preconfigure tasks, the rest of gcc simply depends on these tasks
and have no fetch/unpack/patch tasks of their own.
This means we should get the significant benefits (disk usage/performance) of the
single source tree but in a way which has less potential for problems and is
easier for people to understand. The cost is an extra recipe/some inc files
which is probably a good tradeoff.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
There is a race where:
NOTE: recipe libgcc-initial-4.9.1-r0: task do_configure: Started
NOTE: recipe gcc-runtime-4.9.1-r0: task do_preconfigure: Started
| checking build system type... /home/pokybuild/yocto-autobuilder/yocto-worker/nightly-deb/build/build/tmp/work-shared/gcc-4.9.1-r0/gcc-4.9.1/libgcc/../config.sub: line 1711: syntax error near unexpected token `;;'
| /home/pokybuild/yocto-autobuilder/yocto-worker/nightly-deb/build/build/tmp/work-shared/gcc-4.9.1-r0/gcc-4.9.1/libgcc/../config.sub: line 1711: ` ;;'
| configure: error: /bin/bash /home/pokybuild/yocto-autobuilder/yocto-worker/nightly-deb/build/build/tmp/work-shared/gcc-4.9.1-r0/gcc-4.9.1/libgcc/../config.sub x86_64-linux failed
| WARNING: exit code 2 from a shell command.
so we need to make sure the preconfigure task executes in all shared
work contexts.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Ross Burton <ross.burton@intel.com>
|
|
The gcc-runtime recipe builds the gcc libraries including libstdc++ with
$TARGET_CC_ARCH flags, which include -march=FOO flags that affect
whether atomic instructions are available. This causes an ABI
incompatibility when the compiler by default generates code for less
capable architectures. For example, gcc-runtime libraries on a
Cortex-A8 are built with a different C++11/C++14 mutex implementation
than is used code compiled outside OE and without architecture-specific
flags.
This commit fixes the problem specifically for ABI issues related to
atomic instructions available in ARMV6 and subsequent architectures.
Other ABI incompatibilities may remain in other architectures.
See: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=62100
Signed-off-by: Peter A. Bigot <pab@pabigot.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
musl does not support IBM 128 long double for ppc, instead of
doing complex overrides move it into a pythong snippet which
is easier to read and more compact.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This variable is used to ensure the proper version of --with-float=FOO
is passed to gcc's configure script. gcc also has a --with-fpu=FOO
option that means something different. To avoid confusion, change the
names to be consistent.
Signed-off-by: Peter A. Bigot <pab@pabigot.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Consistent use of whitespace in multi-line assignment, with special
focus on OECONF modifications. Quotes on separate lines, four-space
indentation, one value per line.
Signed-off-by: Peter A. Bigot <pab@pabigot.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The gengtype patch we apply to gcc aims to ensure that the build and host
config headers don't get confused. We're seeing build failures where
both headers have been included, likely due to a race over the configuration
files.
It seems the gengtype-lex.c file isn't being regenerated when it should
and the unconditional inclusion of bconfig.h is resulting in these issues.
The fix is therefore to remove the file, forcing its regeneration.
[YOCTO #6393]
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The do_configure_prepend was duplicated in gcc-4.X.inc and
gcc-configure-common.inc leading to confusion when reading the resulting
do_configure task where the file was processed twice.
The only difference was the removal of the include line for gcc 4.8/4.9.
On mingw were were seeing two issues, firstly that the if statements meant
the values we wanted weren't being set, the second that the include
paths were still wrong as there was no header path set.
To fix the first issue, the #ifdef conditionals were removed, we want
to set these things unconditionally. The second issue is addressed by
setting the NATIVE_SYSTEM_HEADER_DIR variable here (it was already
set in t-oe).
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Within the OE build environment, we supply the correct fpu settings. These
only need to be spelt out for the on-target gcc.
Doing this means the checksums for the core compiler don't depend on the fpu
settings. We exclude the compiler tunes for similar reasons, it doesn't need
to influence the compiler build.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
As far as I can tell this variable is now completely unneeded. It would
only ever get used in target builds and these are now correctly done
in the target environment namespace, not any of our cross environments.
As such, CC and other variables contain the correct compilers and other
tune options and these are correctly picked up when building libgcc,
libstdc++ and others.
I tried to figure out where else these would make any sense and couldn't
find anything. Builds appear fine without them so lets drop the complexity
including the patch adding in this flag to gcc.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This command modifies ${S} and can race against other tasks running do_configure and
having the scripts disappear from under them. To avoid this move to its own
task and work on the shared work directory as a common task.
It needs to be a python task to avoid lots of shell exported variables as
dependencies.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The base_contains is kept as a compatibility method and we ought to
not use it in OE-Core so we can remove it from base metadata in
future.
Signed-off-by: Otavio Salvador <otavio@ossystems.com.br>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
If sed-native is built before these programs, hardcoded paths to sed-native
can end up in scripts and other parts of the system which may cause issues
if they are later used from sstate and sed-native is not installed.
To avoid this, this patch changes the global site configuration to
specify that plain "sed" is fine to be used. We need to spell this
out for gcc since it doesn't see the site files since we don't autoreconf
it. We can remove the values from libtool. We tell perl to use "/bin/sed"
since it requires a path and the system sed should be just fine for it.
[YOCTO #4971]
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
- add a task to setup multilib configuration for target gcc
- this commit adapts Nitin Kamble's work to gcc 4.7
- use a hash for storing arch-dependent multilib options
- patch gcc in order to use the multilib config files from the
build directory
Tests:
root@qemux86-64:~# gcc -m64 t.c -o t
root@qemux86-64:~# file t
t: ELF 64-bit LSB executable, x86-64, version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.16, not stripped
root@qemux86-64:~# ./t
Hello World !
root@qemux86-64:~# gcc -m32 t.c -o t
root@qemux86-64:~# file t
t: ELF 32-bit LSB executable, Intel 80386, version 1 (SYSV), dynamically linked (uses shared libs), for GNU/Linux 2.6.16, not stripped
root@qemux86-64:~# ./t
Hello World !
[YOCTO #1369]
Signed-off-by: Constantin Musca <constantinx.musca@intel.com>
Signed-off-by: Saul Wold <sgw@linux.intel.com>
|
|
This allows BSPs for architectures with no thread support to set (for
example) "GCCTHREADS=no" without having to override all the other configure
parameters.
Signed-off-by: Phil Blundell <pb@pbcl.net>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
with ABI
Signed-off-by: Matthew McClintock <msm@freescale.com>
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
We added the autoconf cache line a while back to ensure that configure doesn't
poke into some hardcoded host paths looking for things it shouldn't. Applying
it as part of do_configure wasn't getting it to the do_compile tasks
where much of the configure scripts are run by gcc. This changes it to a simple
export to ensure it reaches the places it needs to and truly gets rid of
the cross compile badness messages from the logs.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Defaults to n64 ABI
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
Its better than duplicating the overrides
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
this option by default points to /usr/local no matter
what so we cant let it sit on sidelines otherwise it
will access host machine's /usr/local which may not
be desired. So disable this option. This also helps in making
gcc's shared state more consistent
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
When building gcc-cross-canadian libgcc is built using
headers from gcc-crosssdk and not the target sysroot
because we do not pass proper CFLAGS for target bits
so it ends up using CFLAGS that were meant for compiling
canadian gcc itself. It does not show up as a problem
when building SDK with eglibc because eglibc-nativesdk
and eglibc have identical headers. The problem shows
up clearly when you try to build uclibc based meta-toolchain
since then nativesdk libc and target libc are different
Signed-off-by: Khem Raj <raj.khem@gmail.com>
|
|
There were puzzling failures when you make a force recompile of any gcc
component. The error was in do_configure with cross-compilation badness
being detected in config.log files.
gcc is different in that many of the config.log files are generated during
the do_compile phase. This means this host contamination issue has always
been present but only shows up on a rebuild.
The fix is to force the appropriate configuration variable to "none required"
then gcc won't look in the bad locations.
[YOCTO #2279]
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
Using "1" with getVar is bad coding style and "True" is preferred.
This patch is a sed over the meta directory of the form:
sed \
-e 's:\(\.getVar([^,()]*, \)1 *):\1True):g' \
-e 's:\(\.getVarFlag([^,()]*, [^,()]*, \)1 *):\1True):g' \
-i `grep -ril getVar *`
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
We have various variables which are either not quoted at all or are half
quoted. This patch fixes the bad exmaples so everything is consistent.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
The fetch/unpack/patch/headerfix tasks are shared and hence their sstate hashes
should also match. Sadly this is not the case since:
a) gcc-runtime applies an additional patch
b) The do_headerfix task was missing from libgcc
c) The do_headerfix task is a shell task and hence depends
on all exported variables which can vary between cross and target
recipes.
To fix this, the patch moves the patch to the common code, adds
the headerfix task to a common include file and disabled shell
dependencies on the do_headerfix task since its clear in this case
we don't need thsoe dependencies since we just call sed.
With this patch applied, all these recipes now share common sstate checksums.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
This patch introduces a distro feature which enables gcc to produce
both 32bit and 64bit code, and enables binutils to operate on both
32bit and 64bit binaries. It differs from multilib toolchains in
that it does not require to compile a version of the libc for each
architecture variant. However, the code produced for the secondary
architecture will not be linkable against the libc.
v2: - Renamed the feature name from "biarch" to "multiarch". The GCC
installation manual claims that the mips-linux can be made a tri-arch
compiler (http://gcc.gnu.org/install/configure.html)
- For x86_64, the compiler is made bi-arch by default, so nothing
has to be done in particular.
- I analyzed the gcc/config.gcc from GCC sources and added in this
patch all the architectures that could be made biarch with the version
of gcc currently used in OE, which are powerpc, and sparc, in addition
to x86. mips and s390 will probably be supported in future versions of
gcc. For x86 and sparc, only the --enable-targets=all option is valid
to make this work (this option doesn't have any other side effects than
making the compiler bi-arch). For powerpc, I used the
--enable-targets=powerpc64 option (although 'all' also works).
Note: - Untested on powerpc and sparc. But I believe it works the same
as with x86.
- gcc in meta-toolchain is also made multiarch.
Signed-off-by: Julian Pidancet <julian.pidancet@gmail.com>
|
|
This is the result of running the following over the metadata:
sed \
-e 's:bb.data.\(setVar([^,()]*,[^,()]*\), *\([^ )]*\) *):\2.\1):g' \
-e 's:bb.data.\(setVarFlag([^,()]*,[^,()]*,[^,()]*\), *\([^) ]*\) *):\2.\1):g' \
-e 's:bb.data.\(getVar([^,()]*\), *\([^(), ]*\) *,\([^)]*\)):\2.\1,\3):g' \
-e 's:bb.data.\(getVarFlag([^,()]*,[^,()]*\), *\([^(), ]*\) *,\([^)]*\)):\2.\1,\3):g' \
-e 's:bb.data.\(getVarFlag([^,()]*,[^,()]*\), *\([^() ]*\) *):\2.\1):g' \
-e 's:bb.data.\(getVar([^,()]*\), *\([^) ]*\) *):\2.\1):g' \
-i `grep -ril bb.data *`
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|
|
We ensure that:
* the shared work directory contains PR and ensure PR values are consistent across gcc builds
* the regexp to handle library directories is in a specific task and run once
This avoids breakage that was seen in incremental builds after commit
be1f70d68b6b75772ebab8bdff683ddd7c42b0cd where the interpretor could
become corrupted. This was due to the sed expression corrupting
the source directory.
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
|