On Tue, Dec 02, 2025 at 08:46:21PM +0100, Mark Kettenis wrote: > > Date: Tue, 2 Dec 2025 12:52:22 -0600 > > From: Tom Rini <[email protected]> > > > > On Tue, Dec 02, 2025 at 07:49:07PM +0100, Mark Kettenis wrote: > > > > Date: Tue, 2 Dec 2025 12:24:03 -0600 > > > > From: Tom Rini <[email protected]> > > > > > > Hi Tom, > > > > > > > Hey all, > > > > > > > > I am wondering if at this point in time, anyone still builds our host > > > > tools (fw_printenv / fw_getenv, a few others) to run on macOS or > > > > Windows, natively. > > > > > > > > I ask for two reasons. The first of which is that a reason we still have > > > > to support Azure (despite its slowness) for CI is it's where we have > > > > macOS and Windows hosts. But at this point in time there's so many ways > > > > to have Linux userspace running on Windows or macOS that I don't know > > > > that there's any value to these builds. > > > > > > > > The second reason is more macOS specific and is that with: > > > > > > > > commit 8fbcc0e0e839a8e25f636c76e59311033d3817b5 > > > > Author: Marek Vasut <[email protected]> > > > > Date: Thu Nov 13 12:54:51 2025 +0100 > > > > > > > > boot: Assure FDT is always at 8-byte aligned address > > > > > > > > The fitImage may contain FDT at 4-byte aligned address, because > > > > alignment > > > > of DT tags is 4 bytes. However, libfdt and also Linux expects DT to > > > > be at > > > > 8-byte aligned address. Make sure that the DTs embedded in > > > > fitImages are > > > > always used from 8-byte aligned addresses. In case the DT is > > > > decompressed, > > > > make sure the target buffer is 8-byte aligned. In case the DT is > > > > only > > > > loaded, make sure the target buffer is 8-byte aligned too. > > > > > > > > Signed-off-by: Marek Vasut <[email protected]> > > > > > > > > Which does things we must do, we now don't build on macOS. Why? As best > > > > I can tell (and for a general purpose OS, is a good call), memalign(..) > > > > doesn't exist and you need to use posix_memalign, a not drop-in > > > > replacement. We could spend some time reworking the code here for that, > > > > but for now I've instead gone with this workaround so that CI can > > > > continue: > > > > https://lore.kernel.org/u-boot/[email protected]/ > > > > > > Are you sure about this? > > > > > > The code in boot/image-fdt.c is part of U-Boot itself, not the host > > > tools. As such it should be irrelevant whether the host OS provides > > > memalign() or not; U-Boot provides its own implementation. > > > > We use parts of this code as-is (and some other less obvious files) when > > building host tools. > > > > > OpenBSD doesn't have memalign() either. But I can still build the > > > targets I care about on OpenBSD. > > > > > > Or is this about sandbox? Did sandbox ever work on macOS? > > > > It's about "tools-only" build target, but also I suspect that if you try > > the next branch right now it'll also fail and we'll need to figure > > something out. > > Ah, it is only on the next branch. Yes > > In file included from tools/generated/boot/image-fit.c:1: > In file included from ./tools/../boot/image-fit.c:38: > include/malloc.h:824:20: error: conflicting types for 'sbrk' > 824 | extern Void_t* sbrk(ptrdiff_t); > | ^ > /usr/include/unistd.h:449:7: note: previous declaration is here > 449 | void *sbrk(int); > | ^ > 1 error generated. > > Which points out another issue with sharing code between U-Boot itself > and the host tools this way.
Yes, it does. We only kinda-sorta made out OK before this change, and now it's really exposing a few problems. I'd be more inclined to worry about renaming include/malloc.h to include/u-boot/malloc.h (and in turn using <malloc.h> for USE_HOSTCC and <u-boot/malloc.h> normally) to support *BSD than I would be just for macOS / Windows as I do know there are *BSD users out there, along with yourself of course. -- Tom
signature.asc
Description: PGP signature
