In short, I'm confused. We have sys-apps/hwids which provides our pci.ids and usb.ids. sys-apps/hwdata was once (2006) removed and superseded by sys-apps/hwdata-gentoo. Which still contains a blocker against sys-apps/hwdata-redhat, a package dropped only relatively recently in commit 3012cbdf. Now, Plasma 5.19 is going to require "new" package hwdata (which duplicates pci.ids and usb.ids it seems), moved from former fedora-hosted to github - upstream link: https://github.com/vcrhonek/hwdata Specifically, we need pnp.ids for KWin. Is it realistic to include that file in hwids?
So it looks like the source of data for that pnp.ids file is an Excel workbook hosted on download.microsoft.com. https://github.com/vcrhonek/hwdata/blob/master/Makefile#L118 The data seems very static: I doubt any vendors are inventing new ISA PnP ids. I do wonder why KWin needs such obsolete data. Who is running Plasma on an ISA system? We can add a static copy to hwids if needed.
The commit introduced monitor vendor decoding from EDID: https://invent.kde.org/plasma/kwin/-/commit/33a1777a5ab0668da997e18173f4a50643748f1c https://phabricator.kde.org/D10041
Ah ha! According to this Wikipedia page, the UEFI forum is the authority on assigning PnP ids. https://en.wikipedia.org/wiki/Extended_Display_Identification_Data#Structure,_version_1.4 As far as I can tell, the list only seems to be available as an HTML document masquerading as an Excel workbook. https://www.uefi.org/pnp_id_list https://www.uefi.org/uefi-pnp-export I suppose it would be simple enough to write a python script to extract the data.
Earlier this year, hwdata changed its source for pnp.ids to the UEFI table -- see: https://github.com/vcrhonek/hwdata/issues/4 https://github.com/vcrhonek/hwdata/commit/4edeb518d114ee364008b1196474d3faddf51451 So let me provide few points more: - hwdata provides iab.txt, oui.txt, pci.ids, pnp.ids, usb.ids - hwdata is actively maintained (by a Red Hat employee) - hwdata is updated and released every month, usually at the beginning of each month - it is easy to provide feedback to hwdata (see the hwdata issue #4, reported by me), or get fixes there (see the pnp.ids.patch in the repository) - the installed size of hwdata is smaller than the sys-apps/hwids one Hence, IMHO would be a good idea for Gentoo to: a) package hwdata as it is b) drop from sys-apps/hwids what hwdata provides c) make sys-apps/hwids depend on hwdata (to avoid breakages) d) slowly file bugs and transition packages to depend on hwdata directly if they need it It would be much easier to just update hwdata from upstream every month, rather than maintaining a separate project just for some of these files (especially when hwids is not updated regularly...). Especially that there are derivative distros of Gentoo: switching to a regularly updated hwdata will benefit them as well. What do you think? Pino (in this case, with the hat of the Debian maintainer of hwdata)
(In reply to Pino Toscano from comment #4) I think this makes sense. I think the only stuff that hwids provides that hwdata does not is the udev hwdb data. We could probably migrate that back into the udev/systemd packages in Gentoo.
I will plan on working on this soon, once eudev is masked.
Maybe hwdb can be factored out for reuse in eudev? If it can't right now, I'm pretty sure I'll be able to make that work, since I plan to rebase eudev on top of systemd again. Though, I'd say that's in the middle-distant future, because I'm quite busy right now. I can look into factoring it out in a few days.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=8325391d6a8592bf7b3661f8dc47cfc74994c217 commit 8325391d6a8592bf7b3661f8dc47cfc74994c217 Author: Mike Gilbert <floppym@gentoo.org> AuthorDate: 2021-11-24 20:32:10 +0000 Commit: Mike Gilbert <floppym@gentoo.org> CommitDate: 2021-11-28 19:54:00 +0000 sys-apps/hwdata: new package Bug: https://bugs.gentoo.org/717216 Signed-off-by: Mike Gilbert <floppym@gentoo.org> sys-apps/hwdata/Manifest | 1 + sys-apps/hwdata/hwdata-0.353.ebuild | 24 ++++++++++++++++++++++++ sys-apps/hwdata/metadata.xml | 7 +++++++ 3 files changed, 32 insertions(+)
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/proj/kde.git/commit/?id=47b618343c70b76a5b62ce4481c17824a72d4fd2 commit 47b618343c70b76a5b62ce4481c17824a72d4fd2 Author: Andreas Sturmlechner <asturm@gentoo.org> AuthorDate: 2021-11-29 13:19:23 +0000 Commit: Andreas Sturmlechner <asturm@gentoo.org> CommitDate: 2021-11-29 13:20:35 +0000 kde-plasma/kwin: Add missing RDEPEND on sys-apps/hwdata Upstream commit 33a1777a5ab0668da997e18173f4a50643748f1c Bug: https://bugs.gentoo.org/717216 Package-Manager: Portage-3.0.28, Repoman-3.0.3 Signed-off-by: Andreas Sturmlechner <asturm@gentoo.org> kde-plasma/kwin/kwin-5.23.49.9999.ebuild | 7 ++----- kde-plasma/kwin/kwin-9999.ebuild | 7 ++----- 2 files changed, 4 insertions(+), 10 deletions(-)
(In reply to Larry the Git Cow from comment #9) For kwin, I think hwdata should be added to DEPEND as well since the build system looks for the file at build time.
Alternatively, you should pass the path to cmake (hwdata_FOUND, hwdata_DIR, hwdata_PNPIDS_FILE).
(In reply to Mike Gilbert from comment #10) > (In reply to Larry the Git Cow from comment #9) > > For kwin, I think hwdata should be added to DEPEND as well since the build > system looks for the file at build time. It is explicitly marked as runtime dep upstream. > find_package(hwdata) > set_package_properties(hwdata PROPERTIES > TYPE RUNTIME > PURPOSE "Runtime-only dependency needed for mapping monitor hardware vendor IDs to full names" > URL "https://github.com/vcrhonek/hwdata" > )
As a small update on the bug description, sys-apps/hwdata-gentoo was treecleaned in bug 755617. Shall it be closed or do you want to keep it open for the cleanup of sys-apps/hwids?
(In reply to Andreas Sturmlechner from comment #12) > It is explicitly marked as runtime dep upstream. Maybe I misunderstand how cmake works, but doesn't the code in Findhwdata.cmake execute at build time, regardless of how the dependency has been labeled?
That's how it works, but it is of no consequence if missing. Basically a packagers notice, in order to get listed in FeatureSummary when running cmake.
The bug has been referenced in the following commit(s): https://gitweb.gentoo.org/repo/gentoo.git/commit/?id=a86a1f553906b5caac4ead5e176442bf326d75db commit a86a1f553906b5caac4ead5e176442bf326d75db Author: Sam James <sam@gentoo.org> AuthorDate: 2021-12-18 03:35:18 +0000 Commit: Sam James <sam@gentoo.org> CommitDate: 2021-12-18 04:39:25 +0000 sys-fs/eudev: update live ebuild for hwids/hwdb changes upstream eudev upstream now includes updated hwdb files and shouldn't need hwids. Bug: https://bugs.gentoo.org/717216 Signed-off-by: Sam James <sam@gentoo.org> sys-fs/eudev/eudev-9999.ebuild | 14 ++++++++------ 1 file changed, 8 insertions(+), 6 deletions(-)
