Install startfiles for libc variants into the most specific combination
(suffixed sysroot, if applicable + suffixed multi-os dir, if
applicable). Install headers once in every suffixed sysroot (although it
seems that GCC picks up headers from top-level sysroot, GCC manual
claims that sysroot suffix affects headers search path).
In uClibc, this requires a better sanitization of the directory: it
creates symlinks from {sysroot}/usr/lib/{multi_os_dir} to
{sysroot}/lib/{multi_os_dir} and to do so, it counts the number of path
components in the libdir. This breaks if one of such components is `..'
- symlinks contain an extra `../..' then. Since such sanitization had to
be implemented anyway, use it in other places to print more sensible
directory names.
Also, fix the description of configure --host/--target per musl's
configure help message (and its actual code).
Signed-off-by: Alexey Neyman <stilor@att.net>
On some arches (e.g. MIPS) the options like -mabi do not work if
specified more than once (see the comment in 100-gcc.sh). Therefore,
we need to determine which of the options produced by <arch>.sh can
be passed to multilib builds and which must be removed (i.e., which
options vary among the multilibs).
This presents a chicken-and-egg problem. GCC developers, in their
infinite wisdom, do not allow arbitrary multilib specification to be
supplied to GCC's configure. Instead, the target (and sometimes some
extra options) determine the set of multilibs - which may include
different CPUs, different ABIs, different endianness, different FPUs,
different floating-point ABIs, ... That is, we don't know which parts
vary until we build GCC and ask it.
So, the solution implemented here is:
- For multilib builds, start with empty CT_ARCH_TARGET_CFLAGS/LDFLAGS.
- For multilib builds, require core pass 1. Pass 1 does not build any
target binaries, so at that point, our target options have not been
used yet.
- Provide an API to modify the environment variables for the steps that
follow the current one.
- As a part of multilib-related housekeeping, determine the variable
part of multilibs and filter out these options; pass the rest into
CT_TARGET_CFLAGS/LDFLAGS.
This still does not handle extra dependencies between GCC options (like
-ma implying -mcpu=X -mtune=Y, etc.) but I feel that would complicate
matters too much. Let's leave this until there's a compelling case for
it.
Also, query GCC's sysroot suffix for targets that use it (SuperH,
for example) - the default multilib may not work if the command line
specifies the default option explicitly (%sysroot_suffix_spec is not
aware of multilib defaults).
Signed-off-by: Alexey Neyman <stilor@att.net>
For 4 different folders:
${CT_PREFIX_DIR}
${CT_SYSROOT_DIR}
${CT_SYSROOT_DIR}/usr
${CT_PREFIX_DIR}/${CT_TARGET}
.. symlinks from 'lib32' and 'lib64' to 'lib' were created.
This was untidy and incorrect for multilib (the bitness of
the libraries in 'lib32' and 'lib64' will not be the same)
We can not know which folders this toolchain configuration
will require at this time so let them be created on-demand
instead.
Changed by Alexey Neyman: original change removed too much; we
still need to create the default directories because the os
directories are based off them (e.g. `lib/../lib64').
Signed-off-by: Ray Donnelly <mingw.android@gmail.com>
Signed-off-by: Alexey Neyman <stilor@att.net>
GDB's configure mishandles the libexpat.{so,a} libraries when it is
given -static in CFLAGS AND --with-libexpat-prefix in configure's args:
it checks for <prefix>/lib/libexpat.so and finding that, attempts to
link it as `gcc -static .. conftest.c <prefix>/lib/libexpat.so`; this
obviously fails (.so cannot be statically linked), so configure assumes
libexpat is unusable. Thus, --with-libexpat-prefix is dangerous and
should be avoided; instead, configure should find the libraries via the
supplied CC/LD definitions.
If CPATH, C_INCLUDE_PATH, CPLUS_INCLUDE_PATH, or OBJC_INCLUDE_PATH are set, bail out.
These environment variables are known to break crosstool-ng's build.
This closes#327
Signed-off-by: Bryan Hundven <bryanhundven@gmail.com>
We check for apps:
* make
* sed
* grep
* awk
* libtool/libtoolize
* install
* patch
* and more
...during configure. Our scripts should be consistent about using the
variables that define where the found tool was found.
Of course, we do hard-link these tools in buildtools, but that should be
a backup for the components we are building. Our scripts should always
use the tools we find.
Signed-off-by: Bryan Hundven <bryanhundven@gmail.com>
Provides a simpler alternative to editing config to enable
CT_ONLY_DOWNLOAD, doing ct-ng build and then restoring .config.
Signed-off-by: Alexey Neyman <stilor@att.net>
This change updates the CC.* references to CC_GCC.* in the internal
scripts.
Signed-off-by: Ray Donnelly <mingw.android@gmail.com>
Reviewed-by: Bryan Hundven <bryanhundven@gmail.com>
Reviewed-by: Yann Diorcet <diorcetyann@gmail.com>
This commit moves gcc.sh to 100-gcc.sh to accomodate for other
cross-compilers that crosstool-ng might be able to build.
The first, to come soon, is llvm/clang.
Signed-off-by: Ray Donnelly <mingw.android@gmail.com>
Reviewed-by: Bryan Hundven <bryanhundven@gmail.com>
Reviewed-by: Yann Diorcet <diorcetyann@gmail.com>
This commit changes sed, awk, and grep to use the ones we found during
configure time. This helps make the build more consistent.
Signed-off-by: Bryan Hundven <bryanhundven@gmail.com>
BSD grep does not interpret a null alteration. It complains about an
empty sub-expression, e.g.:
$ grep --version && grep -E '^(# |)CT_' .config
grep (BSD grep) 2.5.1-FreeBSD
grep: empty (sub)expression
This patch replaces the null alteration with a zero or once quantifier
which works with both BSD & GNU grep.
$ grep --version && grep -E '^(# )?CT_' .config
grep (BSD grep) 2.5.1-FreeBSD
CT_CONFIGURE_has_xz=y
CT_CONFIGURE_has_svn=y
...
$ ggrep --version && ggrep -E '^(# )?CT_' .config
ggrep (GNU grep) 2.20
Copyright (C) 2014 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>.
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Written by Mike Haertel and others, see
<http://git.sv.gnu.org/cgit/grep.git/tree/AUTHORS>.
CT_CONFIGURE_has_xz=y
CT_CONFIGURE_has_svn=y
...
Signed-off-by: Jason T. Masker <jason@masker.net>
Tested-by: Andreas Bießmann <andreas@biessmann.de>
Signed-off-by: Bryan Hundven <bryanhundven@gmail.com>
Byt the end of the main script, the log file is being moved and
compressed, and the final destination might become read-only at any
time, so we consign stdout/err to oblivion.
This is incorrect, as some actions after may still fail (out of space,
for example).
So, properly restore stdout/err, but also stdin (useless, but harmless)
instead, so the user has a chance to see the error, especially since it
is not logged into the log file.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
The comma is used by the autotools as separator in many sed expressions,
which break if a directory contains commas.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
In case we only download or extract the sources, do not fail while
finishing the toolchain: the test-suite directory may not exist, so
we can't chmod it.
Also, use safer constructs that won't trigger the 'set -e' in case of
failure (eg.: "[ ... ] && ..." is not safe in case the test fails).
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Running as root is really, really dangerous.
Add a runtime-check that refuses to build if running as root.
Can be overriden with a double switch in the menuconfig.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
The current mechanism to check if static linking is possible, and the mesage
displayed on failure, can be puzzling to the unsuspecting user.
Also, the current implementation is not using the existing infrastructure,
and is thus difficult to enhance with new tests.
So, switch to using the standard CT_DoExecLog infra, and use four tests to
check for the host compiler:
- check we can run it
- check it can build a trivial program
- check it can statically link that program
- check if it statically link with libstdc++
That should cover most of the problems. Hopefully.
(At the same time, fix a typo in a comment)
Signed-off-by: Daniel Price <daniel.price@gmail.com>
[yann.morin.1998@free.fr: split original patch for self-contained changes]
[yann.morin.1998@free.fr: use steps to better see gcc's output]
[yann.morin.1998@free.fr: commit log]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Message-Id: <163f86b5216fc08c672a.1353459722@nipigon.dssd.com>
Patchwork-Id: 200536
Signed-off-by: Daniel Price <daniel.price@gmail.com>
[yann.morin.1998@free.fr: split original patch for self-contained changes]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Message-Id: <163f86b5216fc08c672a.1353459722@nipigon.dssd.com>
Patchwork-Id: 200536
Rework binutils in order to provide soon binutils alternative.
Signed-off-by: Yann Diorcet <diorcet.yann@gmail.com>
[yann.morin.1998@free.fr: split up original patch for self-contained changes]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Message-Id: <d3d1d51f399e6d2c1163.1353320546@macbook-smorlat.local>
Patchwork-Id: 199971
On some hosts, and for certain toolchains (eg. toolchain targetting
the upcoming Darwin), it may be necessary to pass arbitrary CFLAGS
and/or LDFLAGS when building the components.
And necessary infrastructure:
- EXTRA_{CFLAGS,LDFLAGS}_FOR_{BUILD,HOST} as config options
- pass those extra flags to components
Fix-up a slight typo in elf2flt at the same time (misnamed cflags).
Signed-off-by: Yann Diorcet <diorcet.yann@gmail.com>
Message-Id: <d24043276c9243a35421.1353077450@macbook-smorlat.local>
Patchwork-Id: 199645
sstrip has been obsoleted for a while now, as it's still broken
for some archs, and there seems to be no incentive to fix it
upstream. Besides, the space gained with sstrip is marginal at
best.
Signed-off-by: Yann Diorcet <diorcet.yann@gmail.com>
Message-Id: <65c8bf534d0647ce52cd.1353320545@macbook-smorlat.local>
Patchwork-Id: 199970
Use the same method as companion tools for providing generic and
extendable companion libs.
Signed-off-by: Yann Diorcet <diorcet.yann@gmail.com>
Message-Id: <515c5c4635d99ebe4877.1353074410@macbook-smorlat.local>
Patchwork-Id: 199613
The extra CFLAGS override the product defaults, causing the product to
be built without optimisation or debug. Be explicit and add these in.
Reported-by: Johannes Stezenbach <js@sig21.net>
Signed-off-by: Michael Hope <michael.hope@linaro.org>
Message-Id: <CANLjY-=3Gbio6nzUPhhevDHV7cUN=6Vigooe9nSf-RnGCqnjog@mail.gmail.com>
Patchwork-Id: 198808
Add an option that, when a command fails:
- starts an interactive shell with the failed command's environment
- attempts re-execution of the failed command, continues, or aborts
at user's whim.
Before starting the debug-shell, the backtrace is printed.
When exiting for an abort, the standard error message is printed.
Based on an idea and a patch from: Johannes Stezenbach <js@sig21.net>
http://sourceware.org/ml/crossgcc/2012-09/msg00144.html
Signed-off-by: Johannes Stezenbach <js@sig21.net>
[yann.morin.1998@free.fr: integrate in the fault handler]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Acked-by: Johannes Stezenbach <js@sig21.net>
Patchwork-Id: 191571
Patchwork-Id: 191668
Avoid error when commands in scripts/crosstool-NG.sh fail
before CT_BUILD_DIR is set.
So we need to remove the backtrace marker of a potential previous
build. Previously, it was implicitly removed because we did remove
the directory it was in, which is no longer the case.
Signed-off-by: Johannes Stezenbach <js@sig21.net>
[yann.morin.1998@free.fr: remove backtrace marker on start of build]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Message-Id: <20121015094615.GA18673@sig21.net>
Patchwork-Id: 191498
It's been a long time the default work-dir changed its name
from 'target' to '.build'.
Change the left-over.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
If either LIBRARY_PATH or LPATH is set, even to the empty string,
the gcc build breaks.
Fix that by bailing-out rather than re-setting.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
These environment variables set search path for gcc at link time, which can break the build.
Signed-off-by: Erik Inge Bolsø <knan-ct@anduin.net>
Message-Id: <alpine.BSF.2.00.1205130131550.21551@anduin.net>
PatchWork-Id: 186872
POSIX 1003.1-2008 does not say whether "set -e" should catch a sub-shell
that exits with !0 (it has a list of conditions to catch, but no list of
conditions not to catch, and this situation is not listed).
bash-3 does not catch such a failure, but bash-4 does. That why, on my
Squeeze system I did not see the issue, while Thomas did on is Lenny chroot.
Reported-by: Thomas Petazzoni <thomas.petazzoni@free-electrons.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@free.fr>
Currently, we rely on an existing external cross-compiler targetting
the target, to build the C library.
This can pause quite a few problems if that compiler is different from
the one we are building, because it could introduce some ABI issues.
This patch removes this dependency, by building the core compilers
as we do for standard cross, and also by building the binutils and
gcc, for running on the build machine.
This means we no longer need to offer the cross-sompiler selection in
the menuconfig.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
It's easier to have as much as possible stuff in the same place to
ease backup/restore, and make things easier to follow.
Move the host companion libraries install dir as a sub-dir of the
build-tools install dir (but not directly in it, it would break
for canadian or cross-native).
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
There really is no good reason to install the core compilers in their
own places, one for each pass. We can install them with the other
build tools.
Also, this implies that:
- there are fewer directories to save/restore
- there are fewer symlinks to create for binutils
- the PATH is shorter
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
The tools found by the new autostuff configure can contain arguments,
for example: grep -E
This needs separating the paths set for the Makfile from the paths
set for the scripts.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Some distributions (eg. openSUSE 12.1) systematically export
the CONFIG_SITE environment variable to point to a custom
script setting misc paths for ./configure.
This can, and does, break when cross-compiling for architectures
that are not supported by this script.
The simple workaround is to unset this variable.
NB: buildroot has a similar fix:
http://git.buildroot.org/buildroot/commit/?id=12c9f7dd6dee9c6029b4f9a12d6aac1516911ab4
Reported-by: Trevor Woerner <twoerner@gmail.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
"${CT_SYSROOT_DIR}/usr/include" is only for "${CT_USE_SYSROOT}" = "y".
We should also mkdir when "${CT_USE_SYSROOT}" != "y".
"${CT_HEADERS_DIR}" can support both cases.
Signed-off-by: Zhenqiang Chen <zhenqiang.chen@linaro.org>
CT_SHELL is undefined.
Thus, the generated wrapper scripts are not executable by the kernel
because they do not contain a valid interpreter.
Use CT_CONFIG_SHELL instead.
Signed-off-by: "Titus von Boxberg" <titus@v9g.de>
CT_EXTRA_FLAGS_FOR_HOST needs a preceding space to separate it from
any other options that have already been set.
Signed-off-by: Michael Hope <michael.hope@linaro.org>
Currently, we check host feature in ./configure. This works only for
cross toolchains, but not for canadian toolchains. ./configure has
absolutely no way to know what the host for the toolchain will be;
only the build scripts know.
So, move the headers & libraries checks from ./configure to the build
scripts, early enough in the build, but not before we know the host
compiler and other tools.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
To avoid variable leakage from one step to another, isolate the
steps from each others by running them in their own sub-shell.
This avoids variables leaking from one step to the others.
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
Allow the user to configure extra flags to pass to the host compiler
at build time. Applies to both C and C++.
Useful on Ubuntu to turn off the stack protector and fortify defaults
so the program stands a better chance of running on other distros.
Signed-off-by: Michael Hope <michael.hope@linaro.org>
[yann.morin.1998@anciens.enib.fr: put the custom flags at the end]
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
When CT_PARALLEL_JOBS is -1, set the number of parallel jobs to the
number of online CPUs + 1. Update documentation to match.
I find this useful when building in the cloud. You can use the same
.config file and have the build adapt to the number of processors
available. Limited testing shows that NCPUS+1 is faster than NCPUS+0
or NCPUS+2.
Signed-off-by: Michael Hope <michael.hope@linaro.org>
The sysroot prefix dir was broken in #4960f5d9f829 due to a mishap
when making the out-of-sysroot lib/ symlink: the './' was mistakenly
changed into a single '.' .
Although Jonathan suggested restoring the missing '/' to restore it to
normal operation, I prefered using an explicit pushd/popd to be extra
sure of the symlink location and target, along with a fix in the sysroot
relative directory calculation.
Reported-by: Jonathan Grundon <JGrundon@xos.com>
Signed-off-by: "Yann E. MORIN" <yann.morin.1998@anciens.enib.fr>
