| Commit message (Collapse) | Author | Age | Files | Lines |
| |
|
| |
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
BuildStream is unable to detect whether output has support for
ANSI color codes automatically when running in a CI and not connected
to a real tty.
It is however typical of CI systems to support the ANSI color escape
codes (but not other terminal escape sequences), for this reason
BuildStream has an explicit --colors option to force colors to be
enabled in scenarios where we cannot detect that it is supported.
This patch causes the CI to use it.
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
| |
|
|
|
|
|
|
| |
GNU gzip was added as qtwebengine depends on the --rsyncable option,
which Busybox doesn't provide. It is more logical to put it in the
same stratum as bzip2 though.
|
|
|
|
|
|
|
|
|
|
| |
See: https://bugs.webkit.org/show_bug.cgi?id=167643
Without this, the build log building with GCC 7 is 1GB in size
due to huge amounts of this kind of thing:
/WebKitGtk.build/Source/JavaScriptCore/b3/B3ArgumentRegValue.cpp:29:1: warning: this use of "defined" may not be portable [-Wexpansion-to-defined]
#if ENABLE(B3_JIT)
|
|
|
|
| |
This isn't actually needed, but it's done now.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
These need to be upgraded together:
- GCC 5.3 fails to build against GLIBC 2.25 due to use of putc() which
triggeres a compile warning. As we seem to build with -Werror this
causes the build to break.
- GLIBC 2.22 fails to build with GCC 7.1 due to int-in-bool-context
errors.
Note that GLIBC 2.22 also fails to build against GLIBC 2.25, failing
with this error:
/tools/include/limits.h:145:17: error: missing binary operator before token "("
#if __GLIBC_USE (IEC_60559_BFP_EXT)
^
|
|
|
|
|
|
|
|
| |
This issue doesn't appear to be fixed upstream, I have done a simple
patch.
The issue is a missing `#include <functional>` causing errors about
`std::function` being undefined. (I'm not sure how this ever worked).
|
|
|
|
|
| |
The upstream looks abandoned so I have applied the patches
to our baserock/ branch.
|
|
|
|
|
|
| |
They helpfully force -Werror on everyone so we have to backport a couple
of `// FALLTHRU` comments to fix the build... the patches don't seem to
even work so I have also disabled -Werror forcefully.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This is to fix build issues with GCC 7 in QtWebKit and QtIvi.
Note that the GENIVI IVI shell patches were merged upstream in QtWayland
commit f144f0f71f1b75df92 so we should no longer need to use a fork of
that component.
qtwebengine 5.9 requires GNU gzip to build -- the `grit` tool (Google
Resource and Internationalization Tool) expects to be able to call `gzip
--rsyncable` which is a feature only provided by GNU gzip. I used
'master' of gzip.git because otherwise the build of the embedded gnulib
components fail with GCC 7 (due to warnings and -Werror).
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This began due to build issues in Clutter with GCC 7, and turned
into a wide-ranging update of GNOME components. Many of these are
needed due to deprecation warnings introduced by the GTK+ upgrade.
The Gettext and GLIB upgrades are needed by GTK+.
The libgdata upgrade was due to deprecation warnings triggered by
GTK+ and this now pulls in autoconf-archive.
Upgrading Clutter requires upgrading Mutter, which in turn requires
new libinput. Note that libinput also required an upgrade to
linux-api-headers (done before). I have updated to Mutter 3.22.3
rather than 3.24 to avoid having to upgrade xrandr.
WebKit has a bug that causes compile failures with GCC 7:
<https://bugs.webkit.org/show_bug.cgi?id=167642>.
The gcrypt upgrade is required to build new WebKit.
gnome-control-center needed to be upgraded which in turn
pulls in network-manager-applet.
|
|
|
|
|
|
|
|
| |
Otherwise you see various link errors like this:
cancel-cancel.o: file not recognized: File truncated
Somehow zero-byte object files are created at random.
|
|
|
|
|
|
| |
Patch from https://github.com/openembedded/meta-openembedded/commit/c9f063c218181e4eac3a0a67d74ce91b898a6443
This patch actually fixes a bug too!
|
|
|
|
| |
Both to fix compile failures with GCC 7
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
Upgrades:
* LLVM (4.0.1)
* libdrm (2.4.81)
* Mesa (17.1.3)
* ragel (6.10)
* fontconfig (2.12.3)
This fixes compile failures in Mesa that occur with GCC 7.1. The libdrm
and LLVM updates are required for latest Mesa.
The LLVM AMDGPU driver is needed by Mesa for the r600 driver. However,
the r600 driver is disabled in Mesa until we integrate libelf
Ragel is updated to fix a build error with GCC 7.
Fontconfig is updated to fix this:
In file included from fcmatch.c:287:0:
fcobjs.h:54:3: error: 'PRI_CHAR_WIDTH_STRONG' undeclared here (not in a function); did you mean 'PRI_WIDTH_STRONG'?
FC_OBJECT (CHAR_WIDTH, FcTypeInteger, NULL)
^~~~~~~~~~~~~~~~~~~~~
PRI_WIDTH_STRONG
fcobjs.h:54:26: error: 'PRI_CHAR_WIDTH_WEAK' undeclared here (not in a function); did you mean 'PRI_WIDTH_WEAK'?
FC_OBJECT (CHAR_WIDTH, FcTypeInteger, NULL)
^~~~~~~~~~~~~~~~~~~
PRI_WIDTH_WEAK
|
|
|
|
|
|
|
|
|
| |
The 0.10.x series is ancient, and seems to have build
failures with GCC 7.
Upstream has also changed for this project, hence the new repo URL.
Some kind of merging happened that meant the new upstream repo isn't
a simple continuation of the previous one.
|
|
|
|
| |
This is needed for LLVM 4.0
|
|
|
|
|
|
|
| |
We are stuck with SYSLINUX 4.06 due to design flaws in how deployment
works with YBD and Morph. In order to fix compile issues with GCC 7 I
have updated the embedded copy of lzo/ in the SYSLINUX source tree
from the syslinux.git 'master' branch.
|
|
|
|
|
|
|
|
|
|
| |
This fixes a compile failure with GCC 7:
kernel/built-in.o: In function `update_wall_time':
(.text+0x69744): undefined reference to `____ilog2_NaN'
Makefile:969: recipe for target 'vmlinux' failed
See also: https://marc.info/?l=linux-kernel&m=147672952517795&w=2
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes a build failure with GCC 7:
amd64-linux-nat.c:497:1: error: conflicting types for 'ps_get_thread_area'
ps_get_thread_area (const struct ps_prochandle *ph,
^~~~~~~~~~~~~~~~~~
In file included from gdb_proc_service.h:25:0,
from amd64-linux-nat.c:50:
/usr/include/proc_service.h:72:17: note: previous declaration of 'ps_get_thread_area' was here
extern ps_err_e ps_get_thread_area (struct ps_prochandle *,
^~~~~~~~~~~~~~~~~~
Makefile:1081: recipe for target 'amd64-linux-nat.o' failed
make[2]: *** [amd64-linux-nat.o] Error 1
|
|
|
|
|
|
|
| |
Mutter 3.22 needs the linux/input-event-codes.h header which was
added fairly recently. This doesn't bump the required kernel version
at all -- that's controlled by the --enable-kernel argument passed
to GLIBC's configure script in strata/build-essential/glibc.morph.
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| |
This fixes a weird issue building 'elfutils' where the new GCC 7
implicit-fallthrough warning would trigger (and fail the build
due to -Werror) even though a suitable /* FALLTHROUGH */ comment
was present in the code.
The issue only occured when building with ccache, not when building with
real GCC. Clearing the ccache didn't help either.
The build works with ccache 3.3.4 so whatever bug caused this must
have now been fixed.
We were previously building from a branch in our mirror of the upstream
Git repo which had files from the release tarball committed. Now we have
proper tarball import repo.
|
| |
|
|
|
|
|
|
|
|
|
|
|
| |
This is how a runner looks like:
$ df -h
Filesystem Size Used Avail Use% Mounted on
/dev/mapper/docker-252:1-262208-36eaa91b86966a7afa39fbdbe717bdec58bc10efc52e09accd3e8e9ee4038658 10G 144M 9.9G 2% /
tmpfs 3.9G 0 3.9G 0% /dev
tmpfs 3.9G 0 3.9G 0% /sys/fs/cgroup
/dev/vda1 79G 1.2G 75G 2% /cache
shm 64M 0 64M 0% /dev/shm
|
|
|
|
|
|
|
|
| |
This was merged but without passing the whole CI; current CI
is broken atm because this change (we need to upgrade several
old chunks first)
This reverts commit 64813d0134a7796f9de9a98ebe2410d97680d51a.
|
|
|
|
| |
The ELF ABI version is different on little-endian.
|
| |
|
|
|
|
|
|
|
|
|
| |
Making linux-api-headers depend on fhs-dirs causes fhs-dirs
to be staged first for anything that depends on build-essential.
This fixes the issue of competing chunks trying to stage symlinks,
or failing to stage the symlinks cause they would otherwise nuke
non-empty directories.
|
|
|
|
|
|
|
|
|
| |
This will make possible to modern distros (with gcc 6) to build
current baserock (which uses gcc 5)
See https://gcc.gnu.org/bugzilla/show_bug.cgi?id=69959
Fixes #8
|
|
|
|
| |
Now there is no need to disable Werror
|
| |
|
| |
|
| |
|
| |
|
| |
|