Changelog:
===========
- don't use O_DIRECT on the RT device
- Fix memory leak
- suggest -x rather than assert for false roots in restore
- fix rootdir due to xfsdump bulkstat misuse
Signed-off-by: Wang Mingyu <wangmy@fujitsu.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Recipes included here are
openldap microsoft-gsl scsirastools pam-ssh-agent-auth sblim-sfcb ckermit nvme-cli xfstests
* these are reproducible only with 32bit MACHINEs and probably related
to time64.inc
http://errors.yoctoproject.org/Errors/Build/183154/
* and microsoft-gsl only with 32bit arm MACHINEs:
http://errors.yoctoproject.org/Errors/Build/183157/
Signed-off-by: Martin Jansa <martin.jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The upstream project now has a version scheme and migrated to meson.
A weak dependency on musl has been upstreamed too.
Adjust the recipe accordingly.
Signed-off-by: Vyacheslav Yurkov <uvv.mail@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
With the UNPACKDIR changes in place the layers are no longer compatible
with the scarthgap release. Drop it from LAYERSERIES_COMPAT and limit
compatibility to styhead only.
Signed-off-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Replace references of WORKDIR with UNPACKDIR where it makes sense to do
so in preparation for changing the default value of UNPACKDIR.
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Technically flatpak and xdg-desktop-portal have a runtime dependency on
fuse3-utils. I was worried that adding it might cause problems with CI since fuse3
is not included in meta-oe.
Given that both packages also require fuse3 at buildtime it could be a good
idea to move fuse3 to avoid a meta-filesystems dependency for meta-oe
Signed-off-by: Markus Volk <f_l_k@t-online.de>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
its needed by fuse3 and right now its only needed by fuse3
in meta-openembedded layers, move it to meta-oe along with
fuse3
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Cc: Markus Volk <f_l_k@t-online.de>
We can make this more widely available as we don't
depend anymore on fuse3.
Signed-off-by: Jose Quaresma <jose.quaresma@foundries.io>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The fuse3 is only used for development purposes and the composefs uses
the kernel overlayfs and erofs on the target.
The composefs-fuse [1] tool is not even installed by default
So all the files installed on the target are the same
with and without the fuse, this can be checked with buildhistory.
[1] f5fed9ff09
Signed-off-by: Jose Quaresma <jose.quaresma@foundries.io>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
For now, the known non-reproducible packages list is stored inside the
autobuilder config.json file. This is not ideal. Let's move this list
into each layers of meta-openembedded.
These lists can be used with, in local.conf:
include conf/include/non-repro-meta-oe.inc
OEQA_REPRODUCIBLE_EXCLUDED_PACKAGES = "${KNOWN_NON_REPRO_META_OE}"
Signed-off-by: Yoann Congal <yoann.congal@smile.fr>
Acked-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This typo would cause all utils tools to be missing in the image when only installing packagegroup-meta-filesystems
Signed-off-by: Yongchang Qiao <qiaoyongchang@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* drop 0003-tests-Makefile-make-sure-group.list-DIRT-exists-befo.patch which
was merged upstream
Signed-off-by: Martin Jansa <martin.jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* add Pending to .patch files where it was accidentally droped
with upgrades or modifications in:
f88e5b146e postgresql: upgrade 15.5 -> 16.2
c904e169db multipath-tools: upgrade 0.9.3 -> 0.9.8
105be9b3d9 unionfs-fuse: upgrade 2.2 --> 3.4
or new patches where the author didn't notice/care:
2a7f74cdb0 dropwatch: Use header files from sysroot instead of build host
f5cc9f272a yasm: improve reproducibility
39028d0d9d python3-pybind11: Restore strip prevention patch
authors of these added to CC, please be more careful with removing
or not adding these or enable patch-status in ERROR_QA for your
builds, see:
https://lists.openembedded.org/g/openembedded-core/topic/104922136#197113
* added with:
for p in `/OE/layers/openembedded-core/scripts/contrib/patchreview.py -v . | grep Missing.Upstream-Status.tag | sed 's/.*(//g;s/)$//g'`; do grep -q ^Upstream-Status: $p || sed -i "s/^---$/\nUpstream-Status: Pending\n---/g" $p; grep -q ^Upstream-Status: $p || sed -i "1iUpstream-Status: Pending\n" $p; done
Signed-off-by: Martin Jansa <martin.jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
Rebased 0005-Replace-off64_t-stat64-with-off_t-stat.patch for 6.6.0.
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* The 20221209 is the commit date of SRCREV since this recipe has no version.
* Add 0001-yaffs_guts.h-define-YTIME_T-if-not-already-defined.patch to fix
build error:
yaffs_guts.h:501:9: error: unknown type name 'YTIME_T'
Signed-off-by: Robert Yang <liezhi.yang@windriver.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
These files are useless for ptests, moreover they cause build QA errors
like
package e2tools-ptest contains reference to TMPDIR
QA Issue: /usr/lib/e2tools/ptest/git/.git/hooks/update.sample contained in package e2tools-ptest requires /usr/bin/sh, but no providers found in RDEPENDS:e2tools-ptest? [file-rdeps]
Signed-off-by: Khem Raj <raj.khem@gmail.com>
This will help in running ptests like core layer, right now there are
just 3 but I hope to get more contributions now that infra is in place
to run all ptests
bitbake meta-filesystems-image-ptest-all
to run a single recipe ptests
bitbake meta-filesystems-image-ptest-fuse3
Signed-off-by: Khem Raj <raj.khem@gmail.com>
python 2 is long unsupported, so we no longer need this variable
Signed-off-by: Tim Orling <tim.orling@konsulko.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
The "update" in commit 2af7d3f63e added a SRCREV rather than updating
SRCREV_xfstests. Change the recipe to actually use SRCREV so that it is
compatible with future AUH updates.
Also replace "/usr" with "${prefix}".
Signed-off-by: Peter Kjellerstedt <peter.kjellerstedt@axis.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* sometimes make install was failing with:
cp: cannot stat 'group.list': No such file or directory
and bunch of non-fatal messages:
mv: failed to preserve ownership for 'group.list': Invalid argument
* this was when tools/mkgroupfile did
mv -f "$new_groups" "$groupfile"
overwritting the group.list file while install-sh was already
copying it to output
* in the end easily reproducible by
1) removing tests/*/group.list before each make install
2) adding some sleep in mkgroupfile before the mv call
Signed-off-by: Martin Jansa <martin.jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
* I'm trying to fix random build failure from world builds:
tests-install sometimes fails as shown bellow:
cp: cannot stat 'group.list': No such file or directory
...
mv: failed to preserve ownership for 'group.list': Invalid argument
whole log:
2023.10.29: http://errors.yoctoproject.org/Errors/Details/752403/
2024.01.14 http://errors.yoctoproject.org/Errors/Details/752404/
* Interestingly I finally triggered this error for first time in my
local build (when rebuilding it after reverting this upgrade -
which might be just coincidence), here is the log:
While looking at this a bit more, I've noticed that install-sh in:
# install-sh script in the project is outdated
# we use the one from the latest libtool to solve installation issues
# It looks like the upstream is not interested in having it fixed :(
# https://www.spinics.net/lists/fstests/msg16981.html
do_configure:prepend() {
cp ${STAGING_DIR_NATIVE}${datadir}/libtool/build-aux/install-sh ${B}
}
is copied only to ${B}, but Makefile still has "cp include/install-sh ."
which was removed in proposed change for upstream:
https://www.spinics.net/lists/fstests/msg16899.html
but here it isn't, so shouldn't we copy it to ${B}/include to make
sure that make won't use the old version?
So I've tried to add:
cp ${STAGING_DIR_NATIVE}${datadir}/libtool/build-aux/install-sh ${B}/include
in do_configure:prepend and then I've triggered the build failure again
this time with 2024.01.14 version as well by running:
for i in `seq -w 1 100`; do bitbake -c cleansstate xfstests; bitbake -k xfstests 2>&1 | tee log.xfstests.$i; done
so the issue isn't fixed in the 2024.01.14 version, but maybe it's
a bug introduced by our install-sh?
It was rejected by upstream in:
https://www.spinics.net/lists/fstests/msg16928.html
This install-sh has been used for performance reason, I'd like keep it
if possible. Would you please describe the failures you hit in
cross-compilation env? It'd be great if we could fix the env issue.
Also the important error isn't:
"mv: failed to preserve ownership for 'group.list': Invalid argument"
because that's shown in all the builds:
xfstests $ grep -c "mv: failed to preserve ownership for 'group.list': Invalid argument" */temp/log.do_install*
2023.10.29-broken/temp/log.do_install:14
2023.10.29-broken/temp/log.do_install.1094046:14
2023.10.29-fixed-after-broken/temp/log.do_install:14
2023.10.29-fixed-after-broken/temp/log.do_install.1094046:14
2023.10.29-fixed-after-broken/temp/log.do_install.1120923:14
2023.10.29/temp/log.do_install:14
2023.10.29/temp/log.do_install.1094046:14
2024.01.14-ok/temp/log.do_install:14
2024.01.14-ok/temp/log.do_install.1186155:14
but the "cp: cannot stat 'group.list': No such file or directory" which is shown only in the ones which in the end failed:
xfstests $ grep cp */temp/log.do_install*
2023.10.29-broken/temp/log.do_install:cp: cannot stat 'group.list': No such file or directory
2023.10.29-broken/temp/log.do_install.1094046:cp: cannot stat 'group.list': No such file or directory
2023.10.29-fixed-after-broken/temp/log.do_install.1094046:cp: cannot stat 'group.list': No such file or directory
2023.10.29/temp/log.do_install:cp: cannot stat 'group.list': No such file or directory
2023.10.29/temp/log.do_install.1094046:cp: cannot stat 'group.list': No such file or directory
So I guess we should just drop this install-sh hack from do_configure completely
as cross-compilation seems to be fixed upstream already in:
https://git.kernel.org/pub/scm/fs/xfs/xfstests-dev.git/commit/?id=5d20084fefaf09e86be26d7c71bd1c7ac2a2b35c
xfstests/2024.01.14-ok/git $ git log --oneline include/install-sh
fa993e51 include/: spdx license conversion
da7916f7 build: Use the original install-sh file from SGI
xfstests/2024.01.14-ok/git $ git log --oneline install-sh
5d20084f build: allow cros-compilation on chromeOS
e3ca126a xfstests: respect DESTDIR when installing
fa6a7868 xfsqa: more sh to bash conversions
491d467f Add GPL license plate to SGI's files.
d259454b Update copyright annotations and license boilerplates to correspond with SGI Legals preferences. Merge of master-melb:xfs-cmds:24329a by kenmcd.
3ef12e38 Update copyright dates (again)
bc4eb461 Undoes mod: xfs-cmds:slinx:120772a Undo xfs-cmds:slinx:120772a, inadvertently whacked a previous mod.
8ce73f86 Update copyright dates
db6944ec need a copy of this here too.
* buildhistory doesn't show any difference, so all the
files are installed in the same location as before
Signed-off-by: Martin Jansa <martin.jansa@gmail.com>
Signed-off-by: Khem Raj <raj.khem@gmail.com>
