Hi Greg and Sasha,
Please consider applying the following commits to 6.1 (they all picked
cleanly for me):
630ae80ea1dd ("tools lib subcmd: Add install target")
77dce6890a2a ("tools lib subcmd: Make install_headers clearer")
5d890591db6b ("tools lib subcmd: Add dependency test to install_headers")
0e43662e61f2 ("tools/resolve_btfids: Use pkg-config to locate libelf")
af03299d8536 ("tools/resolve_btfids: Install subcmd headers")
13e07691a16f ("tools/resolve_btfids: Alter how HOSTCC is forced")
56a2df7615fa ("tools/resolve_btfids: Compile resolve_btfids as host program")
e0975ab92f24 ("tools/resolve_btfids: Tidy HOST_OVERRIDES")
2531ba0e4ae6 ("tools/resolve_btfids: Pass HOSTCFLAGS as EXTRA_CFLAGS to prepare targets")
edd75c802855 ("tools/resolve_btfids: Fix setting HOSTCFLAGS")
The most critical change is 13e07691a16f, which resolves a missing
EXTRA_CFLAGS to the libsubcmd build. Without that EXTRA_CFLAGS, the
Android hermetic toolchain kernel build fails on host distributions
using glibc 2.38 and newer. The majority of those commits are strictly
needed due to dependency/fixes requirements, the few that are not still
seem to be worth bringing in for ease of backporting the rest and do not
appear to cause any problems.
I proposed another solution downstream, which may be more palatable if
people have concerns about this list of changes and the risk of
regressions, but Ian seemed to have some concerns on that thread around
that path and suggested this series of backports instead:
https://android-review.googlesource.com/c/kernel/common/+/2745896
While the number of patches seems large, the final changes are pretty
well self-contained.
Cheers,
Nathan
