On 07.10.2025 10:28, Andrew Cooper wrote: > On 07/10/2025 7:54 am, Roger Pau Monné wrote: >> On Mon, Oct 06, 2025 at 09:08:40PM +0100, Andrew Cooper wrote: >>> --- a/xen/Makefile >>> +++ b/xen/Makefile >>> @@ -6,7 +6,7 @@ this-makefile := $(call lastword,$(MAKEFILE_LIST)) >>> # All other places this is stored (eg. compile.h) should be autogenerated. >>> export XEN_VERSION = 4 >>> export XEN_SUBVERSION = 21 >>> -export XEN_EXTRAVERSION ?= -unstable$(XEN_VENDORVERSION) >>> +export XEN_EXTRAVERSION ?= .0-rc1$(XEN_VENDORVERSION) >> In previous switches to start cutting RC's we didn't add the 1 to the >> XEN_EXTRAVERSION, see 19730dbb3fd8078743d5196bd7fc32f3765557ad for >> example. > > Correct. That was also buggy and complained about. > >> If we do add the rc number we need to remember to bump it >> each time and RC is released. > > Last time we had truly buggy early rc tarballs, the middle ones were > bodged somewhat (incompletely, and not a straight git archive), and by > RC5 we had a commit with the correct XEN_EXTRAVERSION to make the > tarball correct. > > See 4.20.0-rc5
Yet then the question still is whether it is perhaps the tarball making which is what wants adjusting, to not use "$(MAKE) -C xen xenversion", but rather go from, say, the tag in git. Jan
