After merging util/crossgcc: derive date and version from latest commit
(https://review.coreboot.org/c/coreboot/+/30804),
crossgcc build is broken in internal repository due to long version
name;coreboot.org repository is ok because it uses short tag name.
The patch uses "git describe" which is dependent on git tag name.
If tag name is little bit long, it can cause crossgcc build failed.
To avoid this issue, use only short version of hash string
which is fixed length. And it's enough as version string,
because we also use date(CROSSGCC_DATE) together.
TEST=Build crossgcc in both coreboot.org and internal repository
which uses longer tag name and check version string in build log.
Change-Id: I405b2e4e5c05831c25aebf1c73a281adab8ef452
Signed-off-by: Wonkyu Kim <wonkyu.kim@intel.com>
Reviewed-on: https://review.coreboot.org/c/31001
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
This way date and version are automatically updated when util/crossgcc
was changed, the version contains the commit ID and we have less churn
on these variables.
Change-Id: I475ba9578a8bb421d7c342d2569d7de7fcf4161d
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-on: https://review.coreboot.org/c/30804
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: HAOUAS Elyes <ehaouas@noos.fr>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Update to latest version of iasl:
(From the acpica.org changelogs)
* Fixed a regression introduced in version 20180927 that could cause the
compiler to fault, especially with NamePaths containing one or more
carats (^). Such as: ^^_SB_PCI0
* Added a new remark for the Sleep() operator when the sleep time
operand is larger than one second. This is a very long time for the
ASL/BIOS code and may not be what was intended by the ASL writer.
* Implemented detection of extraneous/redundant uses of the Offset()
operator within a Field Unit list. A remark is now issued for these.
For example, the first two of the Offset() operators below are
extraneous. Because both the compiler and the interpreter track the
offsets automatically, these Offsets simply refer to the current
offset and are unnecessary. Note, when optimization is enabled, the
iASL compiler will in fact remove the redundant Offset operators and
will not emit any AML code for them.
Change-Id: I46a1b1be44328aa2172f4741e9fd0c9b0f4e0430
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/c/28944
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
They were not originally printed, and serve no good purpose, so let's
remove them again.
Change-Id: I4e00477f2e143f93fd27ba6a083977a667a3eb48
Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Reviewed-on: https://review.coreboot.org/c/28829
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Otherwise it reduces its expectations on what as and ld take in terms
of arguments, which breaks some edk2 related builds because tons of
-I$path_to_stuff arguments aren't passed along.
Change-Id: I53f87442de03d5ead8a6632d3102d5502065b828
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-on: https://review.coreboot.org/28534
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Martin Roth <martinroth@google.com>
No idea where the escaped parentheses come from but they
are no good. Without this patch I see errors with bash and dash:
./buildgcc: line 1198: (: command not found
./buildgcc: line 1199: (: command not found
The patch uses curly brackets for grouping since they don't
launch a subshell - unlike using unescaped parentheses which
would work too.
shellcheck is happy with either variant (and the original one(!)).
Change-Id: I44fbc659f5b54515e43e85680b1ab0a824b781a7
Signed-off-by: Stefan Tauner <stefan.tauner@gmx.at>
Reviewed-on: https://review.coreboot.org/27771
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-by: Nico Huber <nico.h@gmx.de>
This fixes most of the simpler shellcheck errors in shellcheck 0.4.6.
There are still a few warnings left that weren't simple to fix or
would have required more testing before I was confident in them.
Change-Id: I79ab3614cc1d69d3dfe1e0374e930313f2011cbf
Signed-off-by: Martin Roth <gaumless@gmail.com>
Reviewed-on: https://review.coreboot.org/27598
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
With -D, the newly built toolchain isn't installed into $prefix/...
but into $DESTDIR/$prefix/... while being built for $prefix alone.
This is useful for distributions, but it breaks down when the build
host already has the toolchain installed in $prefix without proper
build isolation (cf. gentoo):
In such cases libgcc etc are built using the new compiler (as gcc's
build system is smart enough to state the path explicitly), but that
compiler then uses its regular algorithm to determine the path to as,
ld, ...
That makes it use the tools from $prefix, which might differ in formats
(assembly, certain object file flags, ...): nds32le-elf in particular
has rather unstable formats still, and so new compilers can't work
with old binutils.
The approach to deal with this is to take an unused path that's
specified by gcc's build system ($out/gcc/$arch/$version) and symlink
it to the new toolchain - these explicitly given directories take
precedence over the default search path, and so the new binutils
are used.
Change-Id: Ia9a262e73f56cd486a2ae07422b598c205a03aed
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Reviewed-on: https://review.coreboot.org/27241
Reviewed-by: Martin Roth <martinroth@google.com>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
We didn't bail out if configuring or building of GCC failed but run
`make install` and later steps instead. This resulted in very confusing
logs that concealed the actual error.
Change-Id: Ia064e0bfd96f0cbad391da3bb19e4dc304d988ff
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/26496
Reviewed-by: Martin Roth <martinroth@google.com>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
That caused the CFLAGS, CXXFLAGS, and '|| touch .failed' to not be taken
into account when building binutils.
Change-Id: I94521eb73cefdc5ed01fbf10122966a54cc28166
Signed-off-by: Vivia Nikolaidou <vivia.nikolaidou@puri.sm>
Reviewed-on: https://review.coreboot.org/25901
Reviewed-by: Youness Alaoui <snifikino@gmail.com>
Reviewed-by: Matt DeVillier <matt.devillier@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Looks like we were unnecessarily dragging this around for some time now.
GCC's installation manual doesn't mention libelf as a requirement and a
build of crossgcc-i386 doesn't show any sign of it being used.
This also fixes a lot issues on non-GNU distributions that were intro-
duced by switching to the elfutils version of libelf.
Change-Id: Iff308a9bed9ae3842557d251b75d1faadfafe0da
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/22773
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Alex Thiessen <alex.thiessen.de+coreboot@gmail.com>
Reviewed-by: Martin Roth <martinroth@google.com>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
In the buildgcc script, there is a check that the tools required are
installed. When a tool is missing, a message is output suggesting an
installation method, e.g. `sudo apt-get install foo` on debian-based
systems.
When run on a true, vanilla debian system, the error message provides
only a generic hint because the `please_install()` function fails to
detect the OS kind. Detection is based on definition of `ID_LIKE` in
`/etc/os-release` yet such systems only define `ID` to `debian`.
This commit closes the detection gap. Tested on debian 9 (stretch).
Change-Id: I3c867837e9157bee13010bd0a005028c369ce55f
Signed-off-by: Alex Thiessen <alex.thiessen.de+coreboot@gmail.com>
Reviewed-on: https://review.coreboot.org/23231
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Add a CBSDK tool set template that can be used in edk2 simply by
appending $prefix/share/edk2config/tools_def.txt to Conf/tools_def.txt.
After that, build -t CBSDK uses the coreboot compilers, providing a more
predictable compiler choice.
Change-Id: I76b38c928b831ee6f31450aa0ad59b4f906f394d
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-on: https://review.coreboot.org/22570
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Also change the tarball from .tar.bz2 to .tar.xz.
Change-Id: I25134dbadf07a2f0cb356c8ac8f2c612a957d176
Signed-off-by: Iru Cai <mytbk920423@gmail.com>
Reviewed-on: https://review.coreboot.org/20806
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
> Could you make in a separate, independent change a update from the
> completely outdated LIBELF (from mr511.de/software/libelf ) to recent
> libelf? Those highly outdated libelf from this unmaintained mr511.de
> webpage should not be used any more since years. There are also a ton
> of security issues like for example: CVE-2017-7607, CVE-2017-7608, ...
> CVE-2017-7613. Recent version of this software is included in the
> elfutils that are available here: https://sourceware.org/elfutils/ ->
> download link:
> https://sourceware.org/elfutils/ftp/0.170/elfutils-0.170.tar.bz2
Remove the obsolete patch, which doesn’t apply anymore, and only
affected the build system, which is different now.
Increment the buildgcc version string as a tool version is changed.
TEST=Running `make crossgcc-i386` succeeds.
Change-Id: Iadd320a18c5d9fe2a82a347e39f01d8b7f8806c2
Signed-off-by: Paul Menzel <paulepanter@users.sourceforge.net>
Signed-off-by: Paul Menzel <pmenzel@molgen.mpg.de>
Reviewed-on: https://review.coreboot.org/21435
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
buildgcc -B (--bootstrap-only) builds only a bootstrap compiler. That
useful if you want to package the cross compilers: first build the
bootstrap compiler, then all required cross compilers in a separate
directory (using the bootstrap compiler through an adjusted PATH).
Change-Id: I089b51d1b898d4cf530845ba51283997fd229451
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-on: https://review.coreboot.org/21683
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Change http to https on many URLs and update llvm.org URLs in buildgcc.
The old URLs are deprecated and now switched to a http forwarder that
can be attacked by MITM attacks.
Change-Id: I68d4fe1a6236ed8540803e11cfc84e44a1d1ca35
Signed-off-by: Doug Gale <doug16k@gmail.com>
Reviewed-on: https://review.coreboot.org/21729
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Reviewed-by: Martin Roth <martinroth@google.com>
This patch implements a relatively simple hash-based verification scheme
for downloaded files (tarballs):
After buildgcc downloads a file or notices that it has already been
downloaded, it hashes the file, and compares the hash against the known
hash stored in util/crossgcc/sum/$filename.cksum. Two errors can occur:
1. The hash file is missing. In this case, crossgcc asks the user to
verify the authenticity of the downloaded file. It also calculates
its hash and stores it in util/crossgcc/sum/$filename.cksum.calc.
If the file is authentic, the user may rename the calculated hash
file to $filename.cksum, so that it can be found the next time
buildgcc is started.
2. The known hash and the calculated hash differ. This is the case that
this patch seeks to protect against, because it may imply that the
downloaded file was unexpectedly changed, either in transit
(Man-in-the-Middle attack) or on the file server that it was
downloaded from. If buildgcc detects such a hash mismatch, it asks
the user to delete the downloaded file and retry, because it can also
be caused by a benign network error. If, however, the error persists,
buildgcc can't continue without risking that the user runs malicious
code, and it stops.
Note: The hash algorithm may be changed in the future, but for now I
left it at SHA-1, to avoid bloating this patch.
Change-Id: I0d5d67b34684d02011a845d00f6f5b6769f43b4f
Signed-off-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
Reviewed-on: https://review.coreboot.org/21592
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Add a missing line-break escape and, rather cosmetic, guard execution of
$CXX which we allow but don't force to be set.
Change-Id: Icf6d3b7de4b7999b8214489f28997964c490d1e9
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/21307
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
This patch has been provided by Mentor Chih-Chyang Chang
on behalf of Andes Technology. It fixes using the coreboot
toolchain to compile the Chrome EC code base on the ITE8320
embedded controller.
The new patch incorporates a fix for the issue previously
fixed by patches/gcc-6.3.0_nds32.patch, so that patch can
be removed.
patches/gcc-6.3.0_riscv.patch needs to be slightly adjusted
to still apply cleanly (configure scripts only).
Change-Id: I0033888360f13ba951b692b3242aab6697ca61b3
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/20901
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Apparently the script's output text is unclear.
Emphasize the file name, keeping the dot to close the sentence.
Change-Id: I1f214b71629eda5fc54e5671ce63e58948343656
Signed-off-by: Idwer Vollering <vidwer@gmail.com>
Reviewed-on: https://review.coreboot.org/21012
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Among other improvements, the new GDB has a new GDB/CLI
command to erase flash memory and a few Python scripting
enhancements.
Change-Id: Ie3852d5bc74617d3e05509e19bbd1caa281da3e0
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/20799
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Martin Roth <martinroth@google.com>
This is sometimes set by packaging systems (eg Gentoo), so give it a
sane preset.
Change-Id: I651fad12128143e8ed5053e7e9871ea271bfc797
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-on: https://review.coreboot.org/20632
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
For distros that package and version gnat independently from gcc (such
as Ubuntu), try to build with gnatgcc first.
This fixes the issue of gcc -print-prog-name=gnat1 failing because gcc
is of a different version.
Change-Id: Icec6d1fba8855e88ac91d47842dcb7f6b9d35461
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-on: https://review.coreboot.org/20517
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
buildgcc was copied to $DEST/share/buildgcc-$VERSION-, missing the
commit id description.
Change-Id: I83d2074b6466b0d99507845dc714a11ab2c58271
Signed-off-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-on: https://review.coreboot.org/20487
Reviewed-by: Kyösti Mälkki <kyosti.malkki@gmail.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Don't ask for bootstrapping in case of a different host GCC major
version. GCC's versioning scheme changed starting with the release
of GCC 5. There are no big changes between the versions any more.
Instead, show the message when the host GCC's version is below 4.
In case GNAT can't be found, ask the user to abort and install it.
Also give hints for Ubuntu where the package versions are a little
messy (e.g. the meta packages gcc and gnat often point to different
versions).
In case GNAT is found but is too old (< 4.9), enable bootstrapping
by default and tell the user that building will take longer.
In all three cases show a timeout to draw the user's attention.
v2: Update GNAT check to also look for `gnatbind`. It has to be
somewhere in $PATH.
Change-Id: I4d9de11d7469e137ede8ad138296d20c0f8ba78f
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/20332
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
"quit" is a signal name. The FreeBSD `sh` interprets
trap quit 1 2 3 15
as command to reset all the respective signal handlers, instead
of setting quit() as handler.
Change-Id: I69b813ab583f15a9dd89a115f7aea66d966f981b
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/20391
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Some environments (<grumble>cros_sdk</grumble>) provide gcc as $CC and
clang++ as $CXX. The latter needs the higher bracket-depth while the
former has no idea what it means, so tell CC and CXX individually.
Change-Id: I72b75fb9bb5df3a9b1561ee8821ec43ada29b24f
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Reviewed-on: https://review.coreboot.org/20365
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Also check for the presence of the given commands or "gcc", "cc" in this
order if $CC is empty. To untangle the given compilers from boostrapped
ones, introduce hostcc() and hostcxx() functions that return the respec-
tive compilers to be used.
Change-Id: Ic947be53eec25331173ac82ed742017ca3fbf83c
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/20331
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
- Change from 'which' to 'command -v'. 'which' is not a posix command.
Change-Id: Icdf18e7e496447157554b8e61b1528f03456536d
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/20230
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
That's what the option is called in the help text. Not
sure where the divergence came from, so let's fix it.
Change-Id: I621aa203da2d314b93de665dbdadbe4a43725375
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/20301
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Philippe Mathieu-Daudé <f4bug@amsat.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Drop Edward's cfe patch because it has been implemented by
upstream clang differently. Instead of
$ clang --print-librt-file-name
the right way to get ahold of the compiler-rt builtin library is
$ clang -rtlib=compiler-rt --print-libgcc-file-name
Change-Id: I8aac5256da5bfb6f7bebeff0959f16b53867c581
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/20274
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Newer versions of clang will need newer versions of CMake (at least
3.4.3) to compile. This patch will enable us to switch to clang 4.0.
Change-Id: I6c91163ce0efd4eb2410cdb433de8be23d510ecd
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/20273
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Philippe Mathieu-Daudé <philippe.mathieu.daude@gmail.com>
Reviewed-by: Arthur Heymans <arthur@aheymans.xyz>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Probably this was never tested as the return to no color "\033[0m"
was printed verbatim.
Change-Id: I7e6e1049b062ffb138ebdaeb62ddc49581ff8db1
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/19811
Tested-by: build bot (Jenkins) <no-reply@coreboot.org>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Ironically enough, libsanitizer is notorious for creating "uninitialized
variable" warnings with different compiler versions than the one it's
shipping with.
Since we don't need it for building the real compiler, just skip it.
Fixes building our compilers using the gnat-gpl 2014 compilers.
Change-Id: I2130dfdf3eaf07d77cd70777419fc0ae4642b843
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Reviewed-on: https://review.coreboot.org/19478
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-by: Philippe Mathieu-Daudé <philippe.mathieu.daude@gmail.com>
We need to rewrite libtool's files (foo.la) a couple of times so it
knows where to look
(while still whining that $DESTDIR$TARGET != $TARGET. well, duh.)
Change-Id: I54cafd47c76d855222ba905b5eb4533a23bdfd34
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Reviewed-on: https://review.coreboot.org/19463
Tested-by: build bot (Jenkins)
Reviewed-by: Philippe Mathieu-Daudé <philippe.mathieu.daude@gmail.com>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Compiling the GNAT frontend of GCC seems to have stabilized since GCC
4.9.0. So build it by default if GNAT >= 4.9 is installed.
TEST=Bootstrapped all GCC versions from 4.9.0 to 6.2 and built the
i386 cross toolchain with each.
Change-Id: I9d1127595dc6b9bcece9c5e5cc7e45f467744ab9
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/18777
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Philippe Mathieu-Daudé <philippe.mathieu.daude@gmail.com>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
We were looking for the wrong file for some time. With bootstrapping
enabled, this resulted in a spurious message about the host GCC being
already built.
Change-Id: Ieb52c5925ea5615c83311319f22693b72f4987f9
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/18776
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
- GCC gets updated from 5.2.0 to 6.3.0:
gcc-6.3.0_riscv.patch is a diff between 5fcb8c4 and 173684b in
riscv-gcc, and it needs gcc-6.3.0_memmodel.patch.
- Binutils goes from 2.26.1 to 2.28:
There is a build error for MIPS gold so I add patch for it.
- GMP gets a bump from 6.1.0 to 6.1.2
- MPFR is updated from 3.1.4 to 3.1.5
- GDB is upgraded from 6.1.1 to 6.1.2
- IASL is changed from 20160831 to 20161222
- LLVM is changed from 3.8.0 to 3.9.1
Change-Id: I20fea838d798c430d8c4d2cc6b07614d967c60c5
Signed-off-by: Iru Cai <mytbk920423@gmail.com>
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/17189
Tested-by: build bot (Jenkins)
There are systems that come with curl but not wget (eg macOS) and they
now have to install one less additional dependency.
Also fix some cosmetic issues in console output and require valid
certificates on https downloads.
Change-Id: Idc2ce892fbb6629aebfe1ae2a95dcef4d5d93aca
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Reviewed-on: https://review.coreboot.org/18048
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-by: Nico Huber <nico.h@gmx.de>
If we use ccache we have to interpret spaces in $CC as separation
characters. The downside is that we can't support spaces in the
compiler's path. But, well...
Change-Id: I4e6e6324389354669a755f570083a40ff00b1bbf
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/18018
Reviewed-by: Martin Roth <martinroth@google.com>
Tested-by: build bot (Jenkins)
CXXFLAGS seems to be used a lot and have to be specified independently
from CFLAGS.
Change-Id: Iff4c76e54a46e908299b532fd848165a3dc04d43
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/17937
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Martin Roth <martinroth@google.com>
GCC 6 can optionally default to building all binaries as position
independent executables (PIE). This breaks linking against static
libraries that are compiled without position independent code (PIC).
Building GMP `--with-pic` in this case seems to be the least fragile
solution.
TEST=Run `make all` and `make BUILDGCC_OPTIONS=-b build-i386` in
util/crossgcc on Debian Stretch.
Change-Id: I5f3185af9c8d599379a628e18724b217b88be974
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/17936
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Jonathan Neuschäfer <j.neuschaefer@gmx.net>
It turned out that newer GNAT versions can build our current (5.3.0)
GNAT without bootstrapping. So adapt the version enforcement.
Change-Id: Ie7189e8bcadeee56cf5c2172e8c0ae7cd534685a
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/17706
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
- Show number of threads being used to build.
- Show the version number of each package when skipping it.
- Show whether the tool is a host or target build.
Change-Id: I1134c08b417a731859e6b25fe38aecf01a85927b
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/17418
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Nico Huber <nico.h@gmx.de>
Otherwise errors similar to "touch: cannot touch
'${TARGETDIR}/.GMP.6.1.0.success': No such file
or directory" might occur.
Change-Id: I4f24c93a25b7d567d3ce14a0415d20fd0778c9c8
Signed-off-by: Patrick Georgi <pgeorgi@chromium.org>
Reviewed-on: https://review.coreboot.org/17603
Tested-by: build bot (Jenkins)
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Previously, the .success file for each target didn't save the version,
of the package that was built. This created problems when someone
wanted to update to a new version and could not rebuild.
Change-Id: I9975b198ac4a7de8ff9323502e1cbd0379a1dbb8
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/17417
Reviewed-by: Nico Huber <nico.h@gmx.de>
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
GCC build instruction recommend to bootstrap a native compiler first.
Not sure, when that is really necessary. A major version change seems
reasonable.
Change-Id: I80a9ec25739b7d33a1d1c7b4b2140d19d89a99ae
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/16675
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Just add some helpers that show parts (major, major.minor) of the GCC
version to be built (buildcc_*) and of the host compiler (hostcc_*).
They will be used in follow-up commits.
Change-Id: I37c12ad1a2d08645f40a9f0f0a479c8d7cc3e127
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/16674
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Tested-by: build bot (Jenkins)
Also remove a dead line that checks for unknown options: We already let
`getopt` check that.
Change-Id: I0e829b266e192757d6e455ee4cc608315bb4b7be
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/16681
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
We accidentally checked the status of `eval` instead.
Change-Id: I1ba258944184ed707ed1f176e528d8266656cb59
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/16680
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Looks like this never worked correctly: There are three argument formats
to GNU getopt and none of them matches what we fed it. The missing
double dash before the `set` arguments proves that we always called it
with parameters that `getopt` did NOT parse.
Change-Id: Ib8343976ef31774b18567a9fc9745a9f58dd287a
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/16679
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Tested-by: build bot (Jenkins)
As we support `getopt` versions that don't know long options, every
option arguments needs a short option.
Also add the long options `--urls` and `--nocolor` to the `getopt`
string.
Change-Id: I11c393c3d90c7a16cdda119594221c85f902ed40
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/16682
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Tested-by: build bot (Jenkins)
Change-Id: I3e3973e1c47505718cf73435156104ab73680441
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/16387
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: Raptor Engineering Automated Test Stand <noreply@raptorengineeringinc.com>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
There are shells where the result of a command substitution is subject
to word splitting (e.g. dash when assigning a value inside an export
statement).
Change-Id: I70a5bc124af7ee621da2bdb4777f3eaba8adafbb
Signed-off-by: Idwer Vollering <vidwer@gmail.com>
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/15820
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: build bot (Jenkins)
The checksum command might appear to be unpredictable only by
checking the OS. Just list the candidates, sorted by possibility.
Change-Id: Ia3f4f5f0f98ff47d322a4f70689cca0bd4fa79fa
Signed-off-by: Zheng Bao <zheng.bao@amd.com>
Signed-off-by: Zheng Bao <fishbaozi@gmail.com>
Reviewed-on: https://review.coreboot.org/11483
Tested-by: build bot (Jenkins)
Reviewed-by: Martin Roth <martinroth@google.com>
Change-Id: I4af90fd2fcfb2a823f9e6b1e975c71581f0b55e9
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/16164
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
On some kind of terms (shell in emacs), the color-ctrl
letters don't work. The backspaces can not delete
correct number of letters. So we don't print color-ctrl
letters in loop.
Change-Id: I1f1729095e8968a9344ed9f1f278f7c78f7110e9
Signed-off-by: Zheng Bao <fishbaozi@gmail.com>
Reviewed-on: https://review.coreboot.org/16066
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Turned out that there are versions of the patch command that use the
left hand side path for new files created by a patch. This behavior is
incompatible with some of our patches. Stripping the topmost dir from
the path with -p1 helps.
While touching that line, I couldn't resist to drop a command
substituion (the `echo $patch`). It really shouldn't be necessary as the
path to the patch file is already expanded in the head of the for loop.
Change-Id: I95398605db6dd54a8b08d8bc84c6602edbea6e10
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/15908
Reviewed-by: Idwer Vollering <vidwer@gmail.com>
Tested-by: build bot (Jenkins)
When no CFLAGS are explicitly provided to it, the GMP configure script
will figure out the best optimization flags to use on its own. In
particular, it will setup the march, mfpu and mtune flags based on
hardware detection.
However, when CFLAGS are provided, they are used as-is and such
detection doesn't happen. When the march, mfpu and mtune flags are not
provided (which happens when GMP wasn't built already), not only will
related optimizations be disabled, but some code might not build because
of missing support. This happens with NEON instructions on ARMv7 hosts.
Thus, it is better not to set CFLAGS and leave it up to the GMP
configure script to get them right and still reuse those later.
Change-Id: I6ffcbac1298523d1b8ddf29a8bca1b00298828a7
Signed-off-by: Paul Kocialkowski <contact@paulk.fr>
Reviewed-on: https://review.coreboot.org/15452
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Tested-by: build bot (Jenkins)
Reviewed-by: Nico Huber <nico.h@gmx.de>
The binutils patch went in without updating the revision,
so we need to update it now. This was done in commit bcfa7ccb
(buildgcc: Update to binutils-2.26.1 & Fix aarch64 build issue)
Change-Id: Ifad4a2e3973f1f60d0ea840945e2bd097e1b4474
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/15712
Tested-by: build bot (Jenkins)
Reviewed-by: Nico Huber <nico.h@gmx.de>
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
- Update to the latest version of GNU binutils
- Add a patch to undo the changes to binutils done by commit c1baaddf
so that arm-trusted-firmware builds correctly again.
Test: Build arm-trusted-firmware (ATF) with this patch. Build ATF
with binutils 2.26.1 changing the '.align x,0' to '.align x', which
changes the padding bytes to NOP instructions. Verify that everything
except the padding bytes is the same.
See https://sourceware.org/bugzilla/show_bug.cgi?id=20364 for more
information about this issue.
Change-Id: I559c863c307b4146f8be8ab44b15c9c606555544
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/15711
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Tested-by: build bot (Jenkins)
Reviewed-by: Furquan Shaikh <furquan@google.com>
Bootstrapping gcc is the recommended way if your host gcc's version
doesn't match the gcc version you're going to build. While a build
with an outdated host gcc usually succeeds, an outdated gnat seems
to be a bigger issue.
v3: Some library controversy: gcc likes the libraries it ships with
most but we don't want to install shared libraries. So we build
them static --disable-shared) and install only the minimum
(libgcc, libada, libstdc++). However, as the code of these
libraries might be used to build a shared library we have to
compile them with `-fPIC`.
v4: o Updated getopt strings.
o The workaround for clang (-fbracket-depth=1024) isn't needed
for bootstrapping and also breaks the build, as clang is only
used for the first stage in that case and gcc doesn't know
that option.
So far build tested with `make BUILDGCC_OPTIONS="-b -l c,ada"` on
o Ubuntu 14.04 "Trusty Tahr" (i386)
o Debian 8 "Jessie" (x86_64) (building python (-S) works too)
o current Arch Linux (x86_64)
o FreeBSD 10.3 (x86_64) (with gcc-aux package)
and with clang host compiler, thus C only: `make BUILDGCC_OPTIONS="-b"`
on
o Debian 8 "Jessie" (x86_64)
o FreeBSD 10.3 (x86_64)
v5: Rebased after toolchain updates to GCC 5.3.0 etc.
Build tested with `make BUILDGCC_OPTIONS="-b -l c,ada"` on
o Debian 8 "Jessie" (x86_64)
Change-Id: Icb47d3e9dbafc55737fbc3ce62a084fb9d5f359a
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/13473
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Refactor build() to make things more flexible:
Add a parameter that tells if we build a package for the host or for a
target architecture. This is just passed to the build_$package()
function and can be used later to take different steps in each case
(e.g. for bootstrapping a host gcc).
Move .success files into the destination directory. That way we can tell
that a package has been built even if the package build directory has
been removed.
Change-Id: I52a7245714a040d11f6e1ac8bdbff8057bb7f0a1
Signed-off-by: Nico Huber <nico.huber@secunet.com>
Reviewed-on: https://review.coreboot.org/13471
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
Update gdb to 7.11 and expat to 2.1.1
riscv64-elf is still broken.
Change-Id: Id7605f4274fcb15f9c3e366f5c492328f70f7956
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/14461
Tested-by: build bot (Jenkins)
Reviewed-by: Idwer Vollering <vidwer@gmail.com>
New tools:
* mpfr 3.1.4
* binutils 2.26
* gcc 5.3.0
* llvm/clang 3.8.0
Patch changes:
* binutils-2.25_fix-aarch64.patch: fixed in 2.26
* binutils-2.25_host-clang.patch: the positions of header file
includes have been adjusted
* binutils-2.25_no-bfd-doc.patch: update to 2.26
* binutils-2.25_riscv.patch: update from riscv-gnu-toolchain
* gcc-5.2.0_elf_biarch.patch: update to 5.3.0
* gcc-5.2.0_gnat.patch: update to 5.3.0
* gcc-5.2.0_libgcc.patch: update to 5.3.0
* gcc-5.2.0_nds32.patch: update to 5.3.0
* gcc-5.2.0_riscv.patch: update from riscv-gnu-toolchain
* cfe-3.7.1.src_frontend.patch: update to 3.8.0
In the latest code of riscv-gnu-toolchain project, the patch for
{binutils,gcc}/config.sub has been removed, and the target is renamed
as riscv32 and riscv64. The `riscv' to `riscv64' change in xcompile is
in another commit.
Test results:
All GCC and LLVM/clang toolchain build successfully.
x86,arm: qemu boots
power8: firmware fails to boot
aarch64,mips: not tested
riscv: firmware fails to build with new binutils
clang: firmware fails to boot
Signed-off-by: Iru Cai <mytbk920423@gmail.com>
Change-Id: I42ce89c29263d768d161c28199994f17d0389633
Reviewed-on: https://review.coreboot.org/14227
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Always set HOSTCFLAGS to the flags GMP was built with, defaulting to
"-Os" if it isn't built yet. Previously, if GMP was already built or
not even in the list of packages to be built, this was silently skipped
and other packages were built with empty HOSTCFLAGS.
Change-Id: I29b2ea75283410a6cea60dc1c92b87573aebfb34
Signed-off-by: Nico Huber <nico.h@gmx.de>
Reviewed-on: https://review.coreboot.org/13550
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
The xz archives are slightly smaller than the bz2 archives for gmp
and mpfr, so use them instead to speed up the download.
Change-Id: I3729455cdbc46e5a0cff119ecca97b0e00c3d402
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/14462
Reviewed-by: Timothy Pearson <tpearson@raptorengineeringinc.com>
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Both packages are not using the target architecture. Drop it,
and remove them from package_uses_targetarch
Change-Id: I58efde4cb7cc39e7e3c31527eb7682e318928100
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/14464
Reviewed-by: Martin Roth <martinroth@google.com>
Tested-by: build bot (Jenkins)
Reviewed-by: Paul Menzel <paulepanter@users.sourceforge.net>
Store both the version number and git hash in the file name
when copying the buildgcc script to the destination directory.
Also, fix the quoting in the lines touched anyways, and move the
script to $TARGETDIR/share/
Change-Id: Ib37dc2be57ee7f0ae18a0b954f537f8b4c2db9d0
Signed-off-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-on: https://review.coreboot.org/14347
Tested-by: build bot (Jenkins)
Reviewed-by: Martin Roth <martinroth@google.com>
Many of the tools and libraries don't use a target architecture, but
they were still getting put in one. This change separates out the
builds that need the target architecture from the ones that don't,
and sets the build directory accordingly.
This will help keep from rebuilding the libraries when building all
of the tools if you keep the temporary files around (-t option).
Change-Id: Id6c17719332f2244657f103f5f07ca7812d51af1
Signed-off-by: Martin Roth <martinroth@google.com>
Reviewed-on: https://review.coreboot.org/14229
Tested-by: build bot (Jenkins)
Reviewed-by: Patrick Georgi <pgeorgi@google.com>
On certain versions of /bin/sh the following sequence
causes problems.
'$CC --version | grep clang &>/dev/null && ...'
The above is a bashish for 2>&1 >/dev/null. However, buildgcc
is interpeted by /bin/sh which doesn't necessarily mean bash.
On dash it's effectively forking grep off into the background
and always evaluating an empty statement to /dev/null while
unconditionally running whatever follows the &&.
Change-Id: Ie3a2ebb12226434d50a7b2a7e254c8b80ae4c46b
Signed-off-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: https://review.coreboot.org/14281
Tested-by: build bot (Jenkins)
Reviewed-by: Martin Roth <martinroth@google.com>
On certain versions of /bin/sh assigning variables with spaces
unquoted leads to failures. Therefore, quote variables that
are known to be passed in that have spaces.
Change-Id: I007c56c3bfb8183bb4b16cf0591f6aa508fd105d
Signed-off-by: Aaron Durbin <adurbin@chromium.org>
Reviewed-on: https://review.coreboot.org/14280
Tested-by: build bot (Jenkins)
Reviewed-by: Stefan Reinauer <stefan.reinauer@coreboot.org>
Reviewed-by: Martin Roth <martinroth@google.com>
Elyes HAOUAS