In a chat some time ago... > [20:24] <arthurzam> asturm: sam_: I wanted to ask your thoughts on > de-stabilize KDE parts on x86, for example www-client/falkon > [20:25] <arthurzam> I suspect too many packages have stable x86 > because it has stable amd64, not because it is relevant > [20:31] <ionen> suspect will have to remove x86 from falkon entirely > at some point, it'll eventually migrate to qtwebengine:6 > which last I know of is unlikely to work on x86 > [20:40] <ionen> not that I've been following this > [...] > [21:59] <asturm> arthurzam: well, what would be the reason (re > falkon)? > [22:00] <asturm> does qtwebengine not build on x86? So, let's clarify here what the problems are and take action afterwards. I suspect the above is about issues with building qtwebengine on *actual* x86 hardware as opposed to amd64 in a 32-bit chroot. I can't help but notice also that x86 stabilisations are delayed when qtwebengine revdeps are involved (see bug 884137, bug 888203). The suggestion in the above IRC snippet was to de-stabilise qtwebengine:5 revdeps. But then I am wondering, if the issue is building on that hardware, what good is it to even keep it in ~x86? dev-qt/qtwebengine:6 has that already settled as chromium dropped ~x86 keyword quite some time ago. So, do we a) de-stabilise dev-qt/qtwebengine:5 +revdeps on x86 b) package.use.mask dev-qt/qtwebengine:5 +revdeps on x86 c) completely drop dev-qt/qtwebengine:5 +revdeps on x86 ?
If I am not mistaken we can basically look at amd64/x32 for what would be necessary: package.use.mask: > app-i18n/fcitx-libpinyin dictionary-manager > dev-lang/idris2 test-full > dev-python/cloudscraper test > games-util/lgogdownloader gui > kde-apps/kde-apps-meta:5 pim > kde-apps/kdenetwork-meta:5 bittorrent telepathy > kde-apps/kleopatra:5 pim > kde-apps/umbrello:5 php > kde-misc/kio-gdrive share > net-irc/quassel urlpreview > net-misc/fatrat bittorrent > net-misc/seafile-client shibboleth > sci-mathematics/yacas gui > sci-physics/root qt5 use.mask: > webengine package.mask: > dev-qt/qtwebengine > dev-qt/qtwebview > dev-python/PyQt6-WebEngine > dev-python/PyQtWebEngine > app-admin/calamares > app-doc/zeal > app-editors/ghostwriter > app-editors/notepadqq > app-editors/retext > app-misc/anki > app-office/kalendar:5 > app-office/kmymoney:5 > app-office/skrooge:5 > app-office/texmaker > app-text/bibletime > app-text/calibre > app-text/cb2bib > app-text/kchmviewer > app-text/sigil > dev-lang/typescript > dev-python/spyder > dev-python/spyder-terminal > dev-python/spyder-unittest > dev-python/spyder-line-profiler > dev-python/spyder-vim > dev-python/spyder-notebook > dev-util/kdevelop:5 > dev-util/kdevelop-php:5 > dev-util/kdevelop-python:5 > gnome-extra/gnome-shell-extension-pop-shell > kde-apps/akonadi-calendar:5 > kde-apps/akonadi-import-wizard:5 > kde-apps/akonadiconsole:5 > kde-apps/akregator:5 > kde-apps/calendarjanitor:5 > kde-apps/calendarsupport:5 > kde-apps/cantor:5 > kde-apps/eventviews:5 > kde-apps/grantlee-editor:5 > kde-apps/incidenceeditor:5 > kde-apps/kaccounts-providers:5 > kde-apps/kaddressbook:5 > kde-apps/kalarm:5 > kde-apps/kalgebra:5 > kde-apps/kdepim-addons:5 > kde-apps/kdepim-meta:5 > kde-apps/kdepim-runtime:5 > kde-apps/kimagemapeditor:5 > kde-apps/kmail:5 > kde-apps/kmail-account-wizard:5 > kde-apps/kmailtransport:5 > kde-apps/knotes:5 > kde-apps/konqueror:5 > kde-apps/konsolekalendar:5 > kde-apps/kontact:5 > kde-apps/korganizer:5 > kde-apps/ktp-accounts-kcm:5 > kde-apps/ktp-text-ui:5 > kde-apps/libksieve:5 > kde-apps/mailcommon:5 > kde-apps/messagelib:5 > kde-apps/mbox-importer:5 > kde-apps/parley:5 > kde-apps/pim-data-exporter:5 > kde-apps/pim-sieve-editor:5 > kde-apps/plasma-telepathy-meta:5 > >=kde-misc/kio-gdrive-22.04.3-r1 > kde-misc/tellico:5 > kde-misc/zanshin:5 > mail-client/kube > media-gfx/digikam:5 > media-gfx/luminance-hdr > media-sound/frescobaldi > media-sound/teamspeak-client:3 > >=media-video/openshot-2.6.0 > media-video/vidify > media-video/vidify-audiosync > net-analyzer/nmapsi > net-libs/signon-ui > net-misc/nextcloud-client > net-p2p/ktorrent:5 > sci-geosciences/qmapshack > sci-mathematics/rkward:5 > sci-visualization/labplot:5 > sys-apps/polychromatic > www-client/falkon > www-client/otter > www-client/qutebrowser > media-video/jellyfin-media-player
I will try to explain my thoughts. I think x86 as a stable arch is long overdue. It seems like we don't have an active userbase of true x86, and I suspect all of them use it as headless, where KDE/Gnome are quite irrelevant. This is a gut feeling and I've no proofs. In the idea that x86 stabilization takes non-trivial time and power, and our only devbox for x86 is for amd64, which has a lot of load (because of stabilizing for amd64), I do prefer to lessen it somewhat by starting to stop stable x86 from a lot of packages. We've been adding x86 for any package that has amd64 for a long time (I think we stopped auto starting with "~amd64 ~x86" only around a year ago) we have a huge amount of stable x86 packages which I suspect we don't really need. I also want to note that looking at "Arch testing dashboard", x86 is at the worst state [1]. (I measure it by amount of pending work, but also the amount we process, as for {arm,ppc}{,64} we process a lot of requests) Which is why I thought to start with the more active project, KDE/Qt. I think we can start by back to ~x86 all the tree depending on qtwebengine, and then maybe full kde & qt. Then we can try to also drop to ~x86 all Gnome/gtk stuff. Do note that it is just my feeling as arch tester, and I'm not the AT lead, council, QA team or anything of that sort - just a helper that raised the idea to break the status-quo. [1] https://dev.gentoo.org/~arthurzam/at_bugs_report.html
Allan Sandfeld Jensen added a comment - 29 Jun '21 10:16 > We had that requirement for Qt5 too, were you using distro patched sources? > > Note, while you could in theory build on a 32bit host, you could only build > the release binaries, and only if you are lucky since linking uses 8-20Gbyte > of memory. Allan Sandfeld Jensen added a comment - 29 Jun '21 10:17 > And note we can still build for x86 arm and mips, just not on x86, arm and > mips. The host needs to be 64bit due to memory requirements while linking.
arthur's comment sums it up quite well. I think we have a lot of debt for x86 where we're stabling things for the sake of it, and some devs are still automatically adding it on new packages (+ stablereqs). We should write something to gentoo-dev about not doing that (unless we did already and I forgot about it). There's two things in conflict for me here: 1. I think not being able to build on $X is a good reason for it to not be marked stable on $X 2. Not being stable on $X means we test it less frequently and are less likely to spot problems. But being honest, if I have time to spend on 32-bit platforms, I'd rather not spend it on something like qtwebengine anyway. Stable-masking it gives us a chance to explain why and also avoids users trying to build it and getting a poor experience. Those who are able to handle binpkgs and set them up should be able to overcome that without being too confused.
(In reply to Sam James from comment #4) > arthur's comment sums it up quite well. I think we have a lot of debt for > x86 where we're stabling things for the sake of it, and some devs are still > automatically adding it on new packages (+ stablereqs). We should write > something to gentoo-dev about not doing that (unless we did already and I > forgot about it). +1
Done.
Re-opening for ~arm. We have 3 build failure bugs (see "see also" in bug 636242) filed against dev-qt/qtwebengine that look quite fundamental. Can we assume no one has really built dev-qt/qtwebengine on ~arm in years, and should we de-keyword it right away?
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=f572a09a0be05cfec6f53c4c4fc34416af768f82 commit f572a09a0be05cfec6f53c4c4fc34416af768f82 Author: Andreas Sturmlechner <asturm@gentoo.org> AuthorDate: 2023-02-07 16:31:48 +0000 Commit: Andreas Sturmlechner <asturm@gentoo.org> CommitDate: 2023-02-07 16:33:01 +0000 profiles: arm: Cleanup obsolete kde-plasma/* webengine package.use.mask Replaced by arm-wide webengine use.mask. Bug: https://bugs.gentoo.org/893234 Signed-off-by: Andreas Sturmlechner <asturm@gentoo.org> profiles/arch/arm/package.use.mask | 9 --------- 1 file changed, 9 deletions(-)
Looks like everything is done here for now.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=515f4ac4feb7922005b4d4de7edb531a2658c12d commit 515f4ac4feb7922005b4d4de7edb531a2658c12d Author: Andreas Sturmlechner <asturm@gentoo.org> AuthorDate: 2024-03-23 12:36:28 +0000 Commit: Andreas Sturmlechner <asturm@gentoo.org> CommitDate: 2024-03-23 16:59:44 +0000 dev-qt/qtwebview: 5.15.13 version bump, unkeywording ~arm/~ppc64 Bug: https://bugs.gentoo.org/893234 Signed-off-by: Andreas Sturmlechner <asturm@gentoo.org> dev-qt/qtwebview/Manifest | 1 + dev-qt/qtwebview/qtwebview-5.15.13.ebuild | 22 ++++++++++++++++++++++ 2 files changed, 23 insertions(+) https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=e50e0290feb058519b96767bc8cf903b6715a60e commit e50e0290feb058519b96767bc8cf903b6715a60e Author: Andreas Sturmlechner <asturm@gentoo.org> AuthorDate: 2024-03-23 12:34:45 +0000 Commit: Andreas Sturmlechner <asturm@gentoo.org> CommitDate: 2024-03-23 16:59:44 +0000 dev-qt/qtwebengine: 5.15.13 version bump, unkeywording ~arm/~ppc64 Bug: https://bugs.gentoo.org/893234 Bug: https://bugs.gentoo.org/924936 Signed-off-by: Andreas Sturmlechner <asturm@gentoo.org> dev-qt/qtwebengine/Manifest | 2 + .../qtwebengine-5.15.13_p20240322.ebuild | 244 +++++++++++++++++++++ 2 files changed, 246 insertions(+)
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/proj/qt.git/commit/?id=7834cf3835c68a65c05a3753dfcd8207feaf6a1a commit 7834cf3835c68a65c05a3753dfcd8207feaf6a1a Author: Andreas Sturmlechner <asturm@gentoo.org> AuthorDate: 2024-05-30 10:28:11 +0000 Commit: Andreas Sturmlechner <asturm@gentoo.org> CommitDate: 2024-05-30 10:28:11 +0000 dev-qt/qtwebview: Unkeyword arm Bug: https://bugs.gentoo.org/893234 Signed-off-by: Andreas Sturmlechner <asturm@gentoo.org> dev-qt/qtwebview/qtwebview-5.15.14.9999.ebuild | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-)
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=c814f3ef18ab542d3757e7f8f244a08b6c6a84dd commit c814f3ef18ab542d3757e7f8f244a08b6c6a84dd Author: Andreas Sturmlechner <asturm@gentoo.org> AuthorDate: 2024-11-22 19:26:41 +0000 Commit: Andreas Sturmlechner <asturm@gentoo.org> CommitDate: 2024-11-23 11:36:04 +0000 dev-qt/qtwebview: unkeyword ~x86 for 5.15, no revdeps Bug: https://bugs.gentoo.org/893234 Signed-off-by: Andreas Sturmlechner <asturm@gentoo.org> dev-qt/qtwebview/qtwebview-5.15.14.ebuild | 2 +- dev-qt/qtwebview/qtwebview-5.15.16.ebuild | 2 +- 2 files changed, 2 insertions(+), 2 deletions(-)
