Date: Mon, 22 Dec 2025 11:02:15 -0500 From: John Baldwin <jhb@FreeBSD.org> To: Rick Macklem <rick.macklem@gmail.com> Cc: src-committers <src-committers@freebsd.org>, dev-commits-doc-all@freebsd.org, doc-committers@freebsd.org Subject: Re: git: 481fce462d - main - Document various missing __FreeBSD_version values. Message-ID: <273f7ba2-f3b0-4a1d-8267-dc48c8d659e2@FreeBSD.org> In-Reply-To: <CAM5tNy7t=6_Enzs8qHVVNFJQE=Af3JOxUtbqCVbUeddaG3N7xA@mail.gmail.com> References: <69447e71.46420.2787bc4a@gitrepo.freebsd.org> <f94be1b6-5441-4930-95f7-c5546664f144@FreeBSD.org> <CAM5tNy7OiEXL6F%2BFyGA7rdwHsimLiATLmUWW=Hnb_n2y0EQK-A@mail.gmail.com> <987ad115-9c14-4f5e-8e1d-ef69f84afd9e@FreeBSD.org> <CAM5tNy7t=6_Enzs8qHVVNFJQE=Af3JOxUtbqCVbUeddaG3N7xA@mail.gmail.com>
index | next in thread | previous in thread | raw e-mail
On 12/18/25 21:56, Rick Macklem wrote: > On Thu, Dec 18, 2025 at 5:45 PM John Baldwin <jhb@freebsd.org> wrote: >> >> On 12/18/25 17:53, Rick Macklem wrote: >>> I'm guilty. But someone said that it should only be bumped >>> if no one else has bumped it recently. (Within a few days?) >>> >>> If it isn't getting bumped for every change, the document >>> is basically useless, isn't it? >> >> The main reason it is documented in the porter's handbook is so that if ports >> need to use conditional logic, this is a reference to which versions to use. >> >> Ports doesn't need perfect granularity (especially on main), just something >> that is "close", hence why bumping more often than, say, once a day is >> probably overkill. > Well, for my case it is inevitably "the internal KAPI between the NFS > modules have changed and all of hem need to be rebuilt from sources". > (Since this cannot affect any port, does it make sense to document it?) > > I do put entries in UPDATING. I do wonder for your case why you are bumping __FreeBSD_version? For other API changes that are internal to the kernel (and modules shipped as part of the kernel), we don't generally bump __FreeBSD_version. Changes to APIs include changes to source files (including headers) so the compiler should already recompile the right things based on dependencies for users who are upgrading via source. Our current binary package systems for the base system (both dist sets and pkgbase) always treat the entire kernel + modules as a unit, so users should never get a mismatched set of NFS-related kernel modules using binary updates either. I think the only case where bumping __FreeBSD_version for these commits can catch something is that if a user builds the NFS kernel modules by hand (e.g. cd /sys/modules/nfsfoo; make) or copies a module manually, the bump will prevent loading a new module while an old kernel is running. It won't though prevent loading an old module against a new kernel (the check we do for kernel modules on main and stable branches is just a <= check, not an exact == check). Given that, I'm not sure you need to be bumping __FreeBSD_version for changes that change APIs between in-tree kernel modules that are not used in ports. -- John Baldwinhome | help
Want to link to this message? Use this
URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?273f7ba2-f3b0-4a1d-8267-dc48c8d659e2>