On Tue, Sep 02, 2025 at 10:57:25PM +0300, Adrian Bunk wrote: > On Fri, Aug 29, 2025 at 10:39:24AM +0200, Helmut Grohne wrote: > > An alternative would be to forbid building on non-x86: > > > > Build-Depends: unsupported-architecture [!any-amd64 !any-i386] > > I would rather suggest > Architecture: amd64 i386 x32 > for two reasons: > > 1. Such Build-Depends tricks are a reasonable workaround for the lack > of ! in the Architecture field, but positive Architecture lists can be > expressed without problem. This is crucial context for understanding developers-reference 5.10.1.9 and .2, which I'd failed at when trying to weigh Helmut's suggestion vs this more obvious-to-me Architecture: list, thanks.
> 2. Hurd also looks quite unsupported.
Testing revealed the supported arches are {linux,kfreebsd}-{i386,amd64},
but not hurd-any (the MAXHOSTNAMELEN compilation error notwithstanding,
hurd glibc advertises iopl() in <sys/io.h>,
but doesn't seem to actually provide a symbol for it).
But we stopped having kfreebsd, so this is moot.
> > It probably works fine on x86 and may not be useful elsewhere.
> "Hardware monitoring without kernel dependencies" - this is the kind
> of software that needs serious porting to every architecture where it
> should run.
Indeed, xmbmon hinges on iopl(2) and x86 I/O ports,
so it's not the spitting image of portability.
Best,
signature.asc
Description: PGP signature
