The bug (being a recipe bug or a bitbake issue) surfaces if you turn do_make_icudata back on in the icu recipe.
In my case I have a need for icu-native to produce the icudata file. Otherwise my native npm/nodejs execution is broken due to the icu.dat file missing (will file this in a different bug).

Reproducer:

clone kirkstone branch

source oe-init-build-env

bitbake icu-native

bitbake -e > icu-native-orig.env
grep BASEHASH icu-native-orig.env > icu-native-orig.basehash


Now:

bbappend icu like this:
-----snip------
PACKAGECONFIG:append = " make-icudata"
do_make_icudata:class-native () {
    ${@bb.utils.contains('PACKAGECONFIG', 'make-icudata', '', 'exit 0', d)}
    cd ${S}
    rm -rf data
    cp -a ${WORKDIR}/data .
    AR='${BUILD_AR}' \
    CC='${BUILD_CC}' \
    CPP='${BUILD_CPP}' \
    CXX='${BUILD_CXX}' \
    RANLIB='${BUILD_RANLIB}' \
    CFLAGS='${BUILD_CFLAGS}' \
    CPPFLAGS='${BUILD_CPPFLAGS}' \
    CXXFLAGS='${BUILD_CXXFLAGS}' \
    LDFLAGS='${BUILD_LDFLAGS}' \
    ICU_DATA_FILTER_FILE=${WORKDIR}/filter.json \
    ./runConfigureICU Linux --with-data-packaging=archive
    oe_runmake
    install -Dm644 ${S}/data/out/icudt${ICU_MAJOR_VER}l.dat ${S}/data/in/icudt${ICU_MAJOR_VER}l.dat
}
-----snip------

IOW enable icudata for icu-native package.

Now run:

bitbake -e > icu-native-new.env
grep BASEHASH icu-native-new.env > icu-native-new.basehash

bitbake icu-native 
NOTE: Tasks Summary: Attempted 120 tasks of which 120 didn't need to be rerun and all succeeded.
# should not build as sstate is considered present


Expected is a rebuild of icu-native with the dat file. This does not happen. 



diff -Napur icu-native-orig.basehash icu-native-new.basehash

--- icu-native-kirkstone-orig.basehash  2023-11-29 20:07:28.316143172 +0000
+++ icu-native-kirkstone-new.basehash      2023-11-29 20:10:17.749386392 +0000
@@ -13,7 +13,7 @@ BB_BASEHASH:task-do_cleansstate="81060e5
 # $BB_BASEHASH:task-do_compile
 BB_BASEHASH:task-do_compile="3b79e1a010a584ddd4430755eee07b526894c29983d763741b85f4f7e386e29e"
 # $BB_BASEHASH:task-do_configure
-BB_BASEHASH:task-do_configure="8093ed634cfadccb2a6ada2aa63f086ad42d39a307a7379a7c3203c0570a09cb"
+BB_BASEHASH:task-do_configure="5a068dd5b5d219753780258a2c5ed4a646bcbcc06d8d3eb49802f9ce43cb5fd9"
 # $BB_BASEHASH:task-do_deploy_source_date_epoch
 BB_BASEHASH:task-do_deploy_source_date_epoch="ccbd1a9eb0adef26e650a567db121e7601d2e7f19819fdad2d5fa48f0b75f2ea"
 # $BB_BASEHASH:task-do_deploy_source_date_epoch_setscene
@@ -27,7 +27,7 @@ BB_BASEHASH:task-do_install="0d74741f27c
 # $BB_BASEHASH:task-do_listtasks
 BB_BASEHASH:task-do_listtasks="12ae738d3561f977d3dc38ff7b21073cd916045f993f8be8e394b5b683ed26c9"
 # $BB_BASEHASH:task-do_make_icudata
-BB_BASEHASH:task-do_make_icudata="8036781e2a98d8eccd7615ae5706cb56245110b494e9d89cb67334ed9a076fb4"
+BB_BASEHASH:task-do_make_icudata="c9231f5124dee9f3afa59194d5986a7e3d4f04a9ec9f6dedd20a9b0edc05ad3e"
 # $BB_BASEHASH:task-do_patch
 BB_BASEHASH:task-do_patch="48ce1f041d3f00fe5af8c7b0d9f0cc023eb2d832c364fa6ada7a17010adce433"
 # $BB_BASEHASH:task-do_populate_lic


It seems the BASEHASHES for setscene are not invalidated through the change made, we pull from sstate. Hmm.



My gut-feeling is that it is an issue with the icu recipe:

https://git.yoctoproject.org/poky/tree/meta/recipes-support/icu/icu_70.1.bb?h=kirkstone#n123

We have an empty 
do_make_icudata() {
    :
}
addtask make_icudata before do_configure after do_patch do_prepare_recipe_sysroot

That is replaced with:
do_make_icudata:class-target () {
    ${@bb.utils.contains('PACKAGECONFIG', 'make-icudata', '', 'exit 0', d)}
    cd ${S}
    rm -rf data
    cp -a ${WORKDIR}/data .
    AR='${BUILD_AR}' \
    CC='${BUILD_CC}' \
    CPP='${BUILD_CPP}' \
    CXX='${BUILD_CXX}' \
    RANLIB='${BUILD_RANLIB}' \
    CFLAGS='${BUILD_CFLAGS}' \
    CPPFLAGS='${BUILD_CPPFLAGS}' \
    CXXFLAGS='${BUILD_CXXFLAGS}' \
    LDFLAGS='${BUILD_LDFLAGS}' \
    ICU_DATA_FILTER_FILE=${WORKDIR}/filter.json \
    ./runConfigureICU Linux --with-data-packaging=archive
    oe_runmake
    install -Dm644 ${S}/data/out/icudt${ICU_MAJOR_VER}l.dat ${S}/data/in/icudt${ICU_MAJOR_VER}l.dat
}

and in my case
do_make_icudata:class-native () {
    ${@bb.utils.contains('PACKAGECONFIG', 'make-icudata', '', 'exit 0', d)}
    cd ${S}
    rm -rf data
    cp -a ${WORKDIR}/data .
    AR='${BUILD_AR}' \
    CC='${BUILD_CC}' \
    CPP='${BUILD_CPP}' \
    CXX='${BUILD_CXX}' \
    RANLIB='${BUILD_RANLIB}' \
    CFLAGS='${BUILD_CFLAGS}' \
    CPPFLAGS='${BUILD_CPPFLAGS}' \
    CXXFLAGS='${BUILD_CXXFLAGS}' \
    LDFLAGS='${BUILD_LDFLAGS}' \
    ICU_DATA_FILTER_FILE=${WORKDIR}/filter.json \
    ./runConfigureICU Linux --with-data-packaging=archive
    oe_runmake
    install -Dm644 ${S}/data/out/icudt${ICU_MAJOR_VER}l.dat ${S}/data/in/icudt${ICU_MAJOR_VER}l.dat
}



The problematic bit might be the empty initial do_make_icudata ?


How would I inspect this on master as there are no BASEHASH'es I can compare ?
Is there a way to produce a similar output on master ?
Maybe its fixed there and only an issue on kirkstone.