On 29 Dec 2025, at 21:58, Mark Johnston <[email protected]> wrote: > > On Mon, Dec 29, 2025 at 12:17:52PM +0000, Dimitry Andric wrote: >> The branch stable/14 has been updated by dim: >> >> URL: >> https://cgit.FreeBSD.org/src/commit/?id=2ade7285253fc5dfe0adfc51f7bd342b9293f253 >> >> commit 2ade7285253fc5dfe0adfc51f7bd342b9293f253 >> Author: Dimitry Andric <[email protected]> >> AuthorDate: 2025-12-26 15:26:13 +0000 >> Commit: Dimitry Andric <[email protected]> >> CommitDate: 2025-12-29 12:16:00 +0000 >> >> tools.build: add sys/cdefs.h to SYSINCS, since lots of other headers use >> it >> >> This is needed to let the legacy stage compile against newer versions of >> sys/font.h, which transitively includes sys/cdefs.h, and requires the >> new __nonstring macro from it. >> >> Fixes: e2c93ed09f25 >> MFC after: 3 days >> >> (cherry picked from commit 1c9ff80f06350fa44b7e50ce09ad665ab0082abb) > > Hi Dimitry, > > After this commit, I can't build stable/14 on main (commit 88b04633c29e > to be exact) anymore. Here's the output from a buildworld, it fails > pretty much immediately: > https://reviews.freebsd.org/P683 > > The same seems to be true on stable/13, but not on stable/15.
Hmm, this turns out to be a bit deeper rabbit hole than I anticipated. The cdefs.h header on stable/14 and stable/13 are quite behind the one on main and stable/15. This causes issues like the above, when __noexcept, __deprecated1 and others are not defined during the legacy stage, and it picks up _some_ headers from the host system, and others from the location installed by tools/build/Makefile. I think I will rollback the MFCs for now, until I can figure out a solution. Ideally we would not have to copy cdefs.h for this bootstrap part, but that leads to other issues. -Dimitry
