summarylogtreecommitdiffstats
AgeCommit message (Collapse)Author
2 daysmerge: master <--- feature/v19.2.2-3Bazaah
* HEAD | \ | * 0ae2952 <github@luxolus.com> pkgbuild: pkgver=19.2.2 pkgrel=3 | / | master Signed-off-by: Bazaah <github@luxolus.com>
3 dayspkgbuild: pkgver=19.2.2 pkgrel=3Bazaah
2025-05-06pkgbuild: pkgver=19.2.2 pkgrel=2Bazaah
2025-05-06pkgbuild: convince ctest to print stuffBazaah
and make sure we don't fail runs on spurious ceph_bcrypt check failures
2025-05-06pkgbuild: use gcc15-zpp-bits.patchBazaah
2025-05-06repo: add gcc15-zpp-bits.patchBazaah
backport of three commit the upstream made a while back for gcc15 Upstream-ref: https://github.com/ceph/ceph/commit/830925f0dd196f920893b1947ae74171a202e825 Upstream-ref: https://github.com/ceph/ceph/commit/9b8d72f15a9b3529d7700aca5aa09d681a1917e1 Upstream-ref: https://github.com/ceph/ceph/commit/3fcd35367283beee3c02b73dc77b7c51028d59dc
2025-05-06repo: add rgw-lua-boost-process-v1.patchBazaah
due to boost::process 1.88 switching the default version of the library to v2, without providing a compatibility layer for the <boost/process> header, we have to manually do this ourselves. Hopefully the upstream can switch to boost/process/v2 when they come across this. References: https://github.com/boostorg/process/issues/480
2025-05-06pkgbuild: use rgw-lua-boost-process-v1.patchBazaah
2025-05-06pkgbuild: use rocksdb-cstdint.patchBazaah
2025-05-06repo: add rocksdb-cstdint.patchBazaah
a few missed <cstdint> imports that gcc15 now complains about
2025-04-22merge: master <--- feature/v19.2.2-1Bazaah
* HEAD | \ | * 294e54e <github@luxolus.com> pkgbuild: pkgver=19.2.2 pkgrel=1 | * 38063e0 <github@luxolus.com> pkgbuild: rere boost-1.87-fixes.patch | * 84435b7 <github@luxolus.com> repo: update boost-1.87-fixes.patch | * 35f0696 <github@luxolus.com> pkgbuild: override cmake minimum version to >=3.13 | * 97eb2de <github@luxolus.com> pkgbuild: split license(s) into three | * f0185e7 <github@luxolus.com> pkgbuild: makedepends += nasm, for ISA erasure-code plugin | / | master Signed-off-by: Bazaah <github@luxolus.com>
2025-04-20pkgbuild: pkgver=19.2.2 pkgrel=1Bazaah
2025-04-04pkgbuild: rere boost-1.87-fixes.patchBazaah
2025-04-04repo: update boost-1.87-fixes.patchBazaah
we update to using a rebased backport of the upstream's fixes for this, as it will make it easier to drop when we move to v20. References: https://github.com/ceph/ceph/pull/61746 Upstream-ref: https://github.com/ceph/ceph/commit/5c1b4703726ec2c63fe4634f8da0a8e2aa7025b0
2025-04-04pkgbuild: override cmake minimum version to >=3.13Bazaah
with the latest update of cmake, _every single vendored dep ceph has fails to build_, because of too-old cmake minvers. Impressive really. I'm not sure if this is dangerous, but 3.13 is ~8 years old now, so hopefully we're fine.
2025-03-12pkgbuild: split license(s) into threeBazaah
this seems to be the way the Archlinux does it, so we'll adopt it
2025-03-12pkgbuild: makedepends += nasm, for ISA erasure-code pluginBazaah
It seems the previous default erasure-code plugin, jerasure has been soft deprecated upstream, and replaced by the "ISA" -- Intel Storage Acceleration -- plugin. One problem: we don't build it currently because of a CMake check: ``` if(HAVE_NASM_X64_AVX2 OR HAVE_ARMV8_SIMD) set(WITH_EC_ISA_PLUGIN TRUE CACHE BOOL "") endif() ``` which I took to mean that I either need AVX2 instructions (or ARM SIMD) on my build server, which I don't have. However, it seems this is not the case, based on the CheckNasm.cmake module in the upstream's private modules, which seems to instead attempt to interrogate nasm about what instruction sets *it* supports, and sets these defines accordingly. References: https://github.com/ceph/ceph/blob/v19.2.1/cmake/modules/CheckNasm.cmake
2025-02-18merge: master <--- feature/v19.2.1-2Bazaah
* HEAD | \ | * c00e90b <github@luxolus.com> pkgbuild: pkgver=19.2.1 pkgrel=2 | * a7d8beb <github@luxolus.com> pkgbuild: use backport-fix-importlib-metadata-compat.patch | * 079099e <github@luxolus.com> repo: add backport-fix-importlib-metadata-compat.patch | * af22ea2 <github@luxolus.com> pkgbuild: use fix-ipv6-support-in-is-addr-in-subnet.patch | * b49de62 <github@luxolus.com> repo: add fix-ipv6-support-in-is-addr-in-subnet.patch | / | master Signed-off-by: Bazaah <github@luxolus.com>
2025-02-17pkgbuild: pkgver=19.2.1 pkgrel=2Bazaah
2025-02-17pkgbuild: use backport-fix-importlib-metadata-compat.patchBazaah
2025-02-17repo: add backport-fix-importlib-metadata-compat.patchBazaah
this fixes a crash in the ceph-volume utility we use for OSDs References: https://github.com/ceph/ceph/pull/59739 Upstream-ref: https://github.com/ceph/ceph/commit/3f1f3450aee368823dc8ae90a8cef13b89b1b950
2025-02-17pkgbuild: use fix-ipv6-support-in-is-addr-in-subnet.patchBazaah
2025-02-17repo: add fix-ipv6-support-in-is-addr-in-subnet.patchBazaah
it seems not many people use ipv6 networks for Ceph clusters. Well, we're one of them and this is needed to not have alerting angry all of the time. This is not yet backported to v19, but exists in the development trunk Upstream-ref: https://github.com/ceph/ceph/commit/1f011aeb549fdeb52d82786b172ab502a971474d References: https://github.com/ceph/ceph/pull/61323
2025-02-13merge: master <--- feature/v19.2.1-1Bazaah
* HEAD | \ | * 3ab57be <github@luxolus.com> pkgbuild: pkgver=19.2.1 pkgrel=1 | * ad7b329 <github@luxolus.com> pkgbuild: bump rbd soname, include extra rgw manpage | * 3eb12be <github@luxolus.com> pkgbuild: disable WITH_QAT stuff harder | * 4e6d7b2 <github@luxolus.com> pkgbuild: use boost-1.87-fixes.patch | * 60377de <github@luxolus.com> repo: add boost-1.87-fixes.patch | * 1170972 <github@luxolus.com> pkgbuild: use quiet-stringop-truncation.patch | * e734306 <github@luxolus.com> repo: add quiet-stringop-truncation.patch | * a818288 <github@luxolus.com> pkgbuild: use fix-cohort-lru-include.patch | * 9842121 <github@luxolus.com> repo: add fix-cohort-lru-include.patch | * 85a796e <github@luxolus.com> pkgbuild: use backport-ceph-exporter.patch | * ff1f8e7 <github@luxolus.com> repo: add backport-ceph-exporter.patch | * 2b9be55 <github@luxolus.com> pkgbuild: use backport-mds-link-boost-urls.patch | * fa36a45 <github@luxolus.com> repo: add backport-mds-link-boost-urls.patch | * 8c5a808 <github@luxolus.com> repo: update boost-1.86-fixes.patch for v19.2.0 | * a690d24 <github@luxolus.com> repo: update test-bluefs-split.patch for v19.2.0 | * 9f173b8 <github@luxolus.com> repo: update fmt10-fixes.patch for v19.2.0 | * b63d268 <github@luxolus.com> repo: update fmt-formatter-const.patch for v19.2.0 | * ed102dc <github@luxolus.com> repo: update py312-fixes.patch for v19.2.0 | * b0310c9 <github@luxolus.com> repo: update disable-mypy-flake8-tests.patch for v19.2.0 | * 6c16e61 <github@luxolus.com> repo: update tox-flake8-git-ls-files.patch for v19.2.0 | * a5a9907 <github@luxolus.com> pkgbuild: rm gcc-14-fixes.patch | * 7f4587e <github@luxolus.com> pkgbuild: rm backport-ceph-volume-unbound-var.patch | * 6206627 <github@luxolus.com> pkgbuild: rm backport-log-runway-expansion.patch | * 77a87e8 <github@luxolus.com> pkgbuild: rm rgw-client-boost-string-view.patch | * 8b8350e <github@luxolus.com> pkgbuild: rm Findfmt.cmake fix | * c00fd1e <github@luxolus.com> pkgbuild: rm pkg.java-cephfs | / | master Signed-off-by: Bazaah <github@luxolus.com>
2025-02-10pkgbuild: pkgver=19.2.1 pkgrel=1Bazaah
2025-02-10pkgbuild: bump rbd soname, include extra rgw manpageBazaah
2025-02-10pkgbuild: disable WITH_QAT stuff harderBazaah
2025-02-10pkgbuild: use boost-1.87-fixes.patchBazaah
2025-02-10repo: add boost-1.87-fixes.patchBazaah
fix a bunch of stuff that boost 1.87 broke around boost::asio, largely involving deprecated function overloads being removed. I did this independently of the upstream, but I realized belatedly that I should have just looked in the PR lists first, as 90% of the fixes already existed, unmerged. References: https://github.com/ceph/ceph/pull/61084
2025-02-08pkgbuild: use quiet-stringop-truncation.patchBazaah
2025-02-08repo: add quiet-stringop-truncation.patchBazaah
this is a minor fix to quiet the gcc line noise from -Wstringop-truncation, despite this being safe in this circumstances because a NULL is appended on the very next line.
2025-02-08pkgbuild: use fix-cohort-lru-include.patchBazaah
2025-02-08repo: add fix-cohort-lru-include.patchBazaah
adds a missing <algorithm> header for std::for_each usage within src/common/cohort_lru.h. This FTBS only seems to trigger during building unit tests (make check)
2025-02-08pkgbuild: use backport-ceph-exporter.patchBazaah
2025-02-08repo: add backport-ceph-exporter.patchBazaah
Seemingly, this was only applied to the v18.x (reef) release, for reasons I can't understand. Nevertheless, we restore it here. This was taken (but heavily edited for v19) from a series of backports for v18.2.4. Upstream-ref: https://github.com/ceph/ceph/commit/bec20dcc8f6f27716d3a4ddffce70cd35eefd8c1
2025-02-08pkgbuild: use backport-mds-link-boost-urls.patchBazaah
2025-02-08repo: add backport-mds-link-boost-urls.patchBazaah
It seems the previous header-only version of Boost.Url no longer is supported, so we need to link it into the mds bin. This partially backports the change from an larger PR, modified to work with 19.2.0 over refs/main References: https://github.com/boostorg/url/blob/boost-1.85.0/include/boost/url/src.hpp References: https://github.com/ceph/ceph/pull/57581 Upstream-ref: https://github.com/ceph/ceph/pull/57581/commits/857eedbe6c9ed80ed0625bd0aa27b1a1e85f8d59
2025-02-08repo: update boost-1.86-fixes.patch for v19.2.0Bazaah
2025-02-08repo: update test-bluefs-split.patch for v19.2.0Bazaah
2025-02-08repo: update fmt10-fixes.patch for v19.2.0Bazaah
2025-02-08repo: update fmt-formatter-const.patch for v19.2.0Bazaah
2025-02-08repo: update py312-fixes.patch for v19.2.0Bazaah
2025-02-08repo: update disable-mypy-flake8-tests.patch for v19.2.0Bazaah
2025-02-08repo: update tox-flake8-git-ls-files.patch for v19.2.0Bazaah
2025-02-06pkgbuild: rm gcc-14-fixes.patchBazaah
the majority of this patch has been applied upstream, and we will reintroduce the dout.h lambda fix in a followup standalone patch
2025-02-06pkgbuild: rm backport-ceph-volume-unbound-var.patchBazaah
fixed upstream Upstream-ref: https://github.com/ceph/ceph/pull/53327
2025-02-06pkgbuild: rm backport-log-runway-expansion.patchBazaah
fixed upstream Upstream-ref: https://github.com/ceph/ceph/pull/50216
2025-02-06pkgbuild: rm rgw-client-boost-string-view.patchBazaah
fixed upstream Upstream-ref: https://github.com/ceph/ceph/pull/51573
2025-02-06pkgbuild: rm Findfmt.cmake fixBazaah
upstream no longer uses this method for finding {fmt}
2025-02-06pkgbuild: rm pkg.java-cephfsBazaah
Nobody has ever used this, and so far as I can tell, it was an experiment that has been abandoned by the upstream; plus the next major version of java will start miscompiling it