https://gcc.gnu.org/bugzilla/show_bug.cgi?id=124066
--- Comment #3 from Richard Biener <rguenth at gcc dot gnu.org> ---
So that does not help. Reason is that while check_cached_effective_target
seems to be set up to have a cache per "target",
proc check_cached_effective_target { prop args } {
global et_cache
set target [current_target_name]
^^^
this does not seem to pick up the effective target from the et-dg-runtest
iteration either.
This all was of course never documented in sourcebuild.texi either.
Also in taget-supports.exp, target_info seems to be some kind of structure
while elsewhere it's a procedure?
Possibly this is all about using --target_board=unix/\{,-m32\}, aka having
multiple multilibs tested?
Some design-direction is much needed here ...
