Bug#928746: unblock: zfs-linux/0.7.13-1
Please start talking to the kernel team now, and let them know your position. If you strongly suspect you're going to file an RC bug in the future, you should be talking now, not just holding back. I'm available to mediate if that ends up being useful.
Bug#928746: unblock: zfs-linux/0.7.13-1
control: retitle -1 unblock: zfs-linux/0.7.12-6 (or 0.7.13-1) control: close -1 Hi Release Team, On 2019-06-03 15:05, Mo Zhou wrote: > Patching the kernel is impossible because kernel maintainers > refused to do that. So that's an invalid solution. After a short discussion with Aron Xu, I realized that I made a mistake about the kernel SIMD problem. It's the ***LTS KERNEL UPDATE*** that breaks ZFS 0.7.12-2. It's not a bug of 0.7.12-2 at all. An LTS KERNEL UPDATE that BREAKS stuff indicates a grave RC. > I'll never argue about ZFS unblock again for Buster. Sorry and I decided to give up the unblock request. We[2] Debian ZoL maintainers decided to do nothing about the kernel SIMD issue and we'll file an RC bug against the kernel immediately after the 10.1 point release because it breaks stuff. [1] https://github.com/zfsonlinux/zfs/commit/e22bfd814960295029ca41c8e116e8d516d3e730 [2] Aron and Me.
Bug#928746: unblock: zfs-linux/0.7.13-1
Hi Paul, On 2019-05-30 19:51, Paul Gevers wrote: > or more severe in Debian BTS terms. I may have been wrong, but then > please point me to the changes so important that you want them in > buster. Please also be prepared to undo the new upstream release and > just fix the bugs that are so important to you. I've already forgotten them and I'm tired to review the diff. Let's forget all the important stuff and only focus on the grave one: Due to a stable kernel change (4.19.37->4.19.38) that makes me sad, we got a grave RC for ZFS 0.7.12-2 (testing) which makes it not suitable for the Buster release: https://bugs.debian.org/cgi-bin/bugreport.cgi?bug=929929 We have two solutions for this stable RC: 1. just unblock 0.7.13-1 . 2. revert debdiff(0.7.12-2,0.7.12-5), then cherry-pick the commits[1] that fix the build issue, then go through t-p-u for 0.7.12-6 . Patching the kernel is impossible because kernel maintainers refused to do that. So that's an invalid solution. > Be aware that requests like this one are draining energy from the > release team. It isn't nice to turn a maintainer down on a request, > repeating the process is worse. Your changes are huge (your explanation > is appreciated), we get several unblock requests per day and we have a > freeze policy in place to manage it. Please don't push your pet project > so hard if it doesn't meet the policy that you are driving the > volunteers in the release team away. Thanks for also considering our time. Sorry for that. My motivation is just that 0.7.13-1 is really expected to reduce the number of problems compared to 0.7.12-2... Please make a choice between the two proposed solution above. If release team doesn't trust every line of code, please just choose 2. I'll never argue about ZFS unblock again for Buster. [1] Patch: https://github.com/zfsonlinux/zfs/commit/e22bfd814960295029ca41c8e116e8d516d3e730
Bug#928746: unblock: zfs-linux/0.7.13-1
On 30 mai 2019 21:51, Paul Gevers wrote: > Control: tags -1 moreinfo > > Hi Mo, > > On Thu, 09 May 2019 22:09:18 -0700 Mo Zhou wrote: >> zfs-linux (= 0.7.13-1) is 66 days in unstable and there is no new bug >> for it. >> Compared to (0.7.12-2), the (0.7.13-1) version in unstable only >> introduces >> bug fixes. Aron has already applied for an unblock but it was rejected. >> Here I'm requesting for unblock again. > > I checked bug #923770 again (it was filed by you by the way). As I said > in that bug, I didn't spot anything that was at the level of important > or more severe in Debian BTS terms. I may have been wrong, but then > please point me to the changes so important that you want them in > buster. Please also be prepared to undo the new upstream release and > just fix the bugs that are so important to you. I see a problem with longterm kernel is Buster (4.19). 4.19.38 doesn't build with zfs 0.7.12 OK, 4.19.37 will be the Buster kernel, but has we upgrade recent kernel from longterm branch for stable release the next stable release will probably break zfs. So zfs 0.7.13 is probably a good choice for Buster. This is because kernel 5.0 "drops export of __kernel_fpu_begin/end" has been back-ported to 4.19.38 (and also 4.14.120) kernel and fixed in zfs 0.7.13 https://github.com/NixOS/nixpkgs/issues/60907 https://cdn.kernel.org/pub/linux/kernel/v4.x/ChangeLog-4.19.38 https://cdn.kernel.org/pub/linux/kernel/v4.x/ChangeLog-4.14.120 Christian
Bug#928746: unblock: zfs-linux/0.7.13-1
Control: tags -1 moreinfo Hi Mo, On Thu, 09 May 2019 22:09:18 -0700 Mo Zhou wrote: > zfs-linux (= 0.7.13-1) is 66 days in unstable and there is no new bug > for it. > Compared to (0.7.12-2), the (0.7.13-1) version in unstable only > introduces > bug fixes. Aron has already applied for an unblock but it was rejected. > Here I'm requesting for unblock again. I checked bug #923770 again (it was filed by you by the way). As I said in that bug, I didn't spot anything that was at the level of important or more severe in Debian BTS terms. I may have been wrong, but then please point me to the changes so important that you want them in buster. Please also be prepared to undo the new upstream release and just fix the bugs that are so important to you. Be aware that requests like this one are draining energy from the release team. It isn't nice to turn a maintainer down on a request, repeating the process is worse. Your changes are huge (your explanation is appreciated), we get several unblock requests per day and we have a freeze policy in place to manage it. Please don't push your pet project so hard if it doesn't meet the policy that you are driving the volunteers in the release team away. Thanks for also considering our time. Paul signature.asc Description: OpenPGP digital signature
Bug#928746: unblock: zfs-linux/0.7.13-1
Package: release.debian.org Severity: normal User: release.debian@packages.debian.org Usertags: unblock X-Debbugs-CC: a...@debian.org Please unblock package zfs-linux zfs-linux (= 0.7.13-1) is 66 days in unstable and there is no new bug for it. Compared to (0.7.12-2), the (0.7.13-1) version in unstable only introduces bug fixes. Aron has already applied for an unblock but it was rejected. Here I'm requesting for unblock again. The difference between upstream 0.7.12 and 0.7.13 includes linux 5.0 compat fixes, and a series of bug fixes. I can revert the linux 5.0 compat fixes if you think that's irrelated to the Buster release with the 0.7.13-2 upload. I can also revert some enhancement patches if release team doesn't like them. Even if we reverted the linux 5.0 compat and some enhancements, it is still better than letting 0.7.12-2 stay in Buster, because in that way we still have bug fixes. On the debian packaging side, the difference between 0.7.12-2 (testing) and 0.7.13-1 (unstable) includes manually cherry-picked patches (by Colin Ian King(ubuntu diff), Aron Xu, and Me), fixes for the problematic autopkgtest script I wrote, and a small update for the Debian+OpenRC support patch. Most of the newly added patches were already shipped by the ubuntu package more than 3 months ago. zfs (0.7.12-5) was uploaded to unstable before (hardfreeze - 12 days). zfs (0.7.13-1) was uploaded to unstable on (hardfreeze - 8 days). I should have waited for some time for the 0.7.12-5 migration... Anyway let's see the diffstat and I'll comment every bit of change in detail. Please at least accept some bug fixes for Buster. And tell me which of the enhancements (incl. linux 5.0 compat) are not acceptible so that I can revert them. (explain the reason for the unblock here) ``` ~/D/z/zfs ❯❯❯ git diff debian/0.7.12-2 debian/0.7.13-1 --stat | cat META | 2 +- Makefile.in| 2 + aclocal.m4 | 2 + cmd/Makefile.in| 2 + cmd/arc_summary/Makefile.in| 2 + cmd/arcstat/Makefile.in| 2 + cmd/dbufstat/Makefile.in | 2 + cmd/fsck_zfs/Makefile.in | 2 + cmd/mount_zfs/Makefile.am | 2 + cmd/mount_zfs/Makefile.in | 4 + cmd/raidz_test/Makefile.in | 2 + cmd/vdev_id/Makefile.in| 2 + [linux 5.0 compat] *.am and *.in was updated for linux 5.0 cmd/vdev_id/vdev_id| 209 - [enhancements] 2b8c3cb0c83681d7425fa5bf567e726b7ba975e9 vdev_id: extension for new scsi topology 41f7723e9cb260f313f99d667142e37617812fca vdev_id: new slot type ses c32c2f17d06cea5dc298b404fad7696921e490e0 Add enclosure_symlinks option to vdev_id cmd/zdb/Makefile.in| 2 + cmd/zed/Makefile.in| 2 + cmd/zfs/Makefile.in| 2 + cmd/zgenhostid/Makefile.in | 2 + cmd/zhack/Makefile.in | 2 + cmd/zinject/Makefile.in| 2 + cmd/zpios/Makefile.in | 2 + cmd/zpool/Makefile.in | 2 + cmd/zstreamdump/Makefile.in| 2 + cmd/ztest/Makefile.in | 2 + [linux 5.0 compat] all the *.in files above cmd/ztest/ztest.c | 4 +- [trivial C code change] - .zo_pool = { 'z', 't', 'e', 's', 't', '\0' }, - .zo_dir = { '/', 't', 'm', 'p', '\0' }, + .zo_pool = "ztest", + .zo_dir = "/tmp", cmd/zvol_id/Makefile.in| 2 + config/kernel-access-ok-type.m4| 21 + config/kernel-bio_set_dev.m4 | 35 +- config/kernel-fpu.m4 | 41 +- config/kernel-misc-minor.m4| 26 + config/kernel.m4 | 4 +- [linux 5.0 compat] the above several files were changed due to linux compat fix configure | 599 ++- Mostly due to newly added kernel feature checks. Deleted lines are due to version number bump. configure.ac | 3 + contrib/Makefile.in| 2 + contrib/bash_completion.d/Makefile.in | 2 + contrib/dracut/02zfsexpandknowledge/Makefile.in| 2 + contrib/dracut/90zfs/Makefile.am | 1 + contrib/dracut/90zfs/Makefile.in | 3 + contrib/dracut/90zfs/module-setup.sh.in| 4 +- contrib/dracut/Makefile.in