diff options
author | Richard Purdie <richard.purdie@linuxfoundation.org> | 2021-08-01 14:04:15 +0100 |
---|---|---|
committer | Richard Purdie <richard.purdie@linuxfoundation.org> | 2021-08-02 15:44:11 +0100 |
commit | 825043ec771aad54c055ec30aee007a793183487 (patch) | |
tree | 69f6058a0b0798159bbff494a3f183f98d132899 /meta/classes/packagedata.bbclass | |
parent | ccd83269b7cbe27270c10c7a4fa0ea5498a1a624 (diff) | |
download | poky-825043ec771aad54c055ec30aee007a793183487.tar.gz |
populate_sdk_ext: Fix handling of TOOLCHAIN_HOST_TASK in the eSDK case
The current way TOOLCHAIN_HOST_TASK is handled within eSDK builds is
causing much confusion, even to people who should understand the code.
For a normal SDK, some layers append to TOOLCHAIN_HOST_TASK unconditionally
which is fine, until the eSDK tries to override the variable to it's own
values. It does not support or use packages installed using this variable
and would use native recipes instead, it is a very different approach.
In the referenced bug, binaries are added but not relocated leading to
confusing errors.
To fix this, add a new variables for the eSDK TOOLCHAIN_HOST_TASK
to be explict and force the eSDK code to use this instead. The setVar
in non-parsing context will clear out any appends resolving the current
issues. The patch also gets rid of some dubious task override use.
[YCOTO #14047]
(From OE-Core rev: 30912ba084aa8600156edddbe4f2db0b85e869d0)
Signed-off-by: Richard Purdie <richard.purdie@linuxfoundation.org>
Diffstat (limited to 'meta/classes/packagedata.bbclass')
0 files changed, 0 insertions, 0 deletions