image.bbclass: exclude deploy-${PN}-image-complete from rm_work
PATCH 1/1 image.bbclass: exclude deploy-${PN}-image-complete from rm_work
From: Ming Liu <[email protected]>
The results in ${IMGDEPLOYDIR} generated by a image task could be used by later image tasks, hence they should be excluded by rm_work. Or else when a image task gets rerun but one of its dependent image task does not, then the results generated by this dependent image task would not be there in ${IMGDEPLOYDIR}.
--- meta/classes/image.bbclass | 4 ++++ 1 file changed, 4 insertions(+)
meta/classes/image.bbclass +4 -0
@@ -84,6 +84,10 @@ PACKAGE_INSTALL_ATTEMPTONLY ?= "${FEATURE_INSTALL_OPTIONAL}" IMGDEPLOYDIR = "${WORKDIR}/deploy-${PN}-image-complete" +# The results in ${IMGDEPLOYDIR} generated by a image task could be used by +# later image tasks, hence they should be excluded by rm_work. +RM_WORK_EXCLUDE_ITEMS += "deploy-${PN}-image-complete" + # Images are generally built explicitly, do not need to be part of world. EXCLUDE_FROM_WORLD = "1"
Signature · 1 line
2.7.4
Quoted 4 lines · On Fri, 2019-08-02 at 17:12 +0800, [email protected] wrote:
From: Ming Liu <[email protected]>
The results in ${IMGDEPLOYDIR} generated by a image task could be used by later image tasks.
No they shouldn't.
Quoted 4 lines
, hence they should be excluded by
rm_work. Or else when a image task gets rerun but one of its dependent image task does not, then the results generated by this dependent image task would not be there in ${IMGDEPLOYDIR}.
Nothing should be poking into IMGDEPLOYDIR outside of the image tasks.
do_image_complete[sstate-inputdirs] = "${IMGDEPLOYDIR}" do_image_complete[sstate-outputdirs] = "${DEPLOY_DIR_IMAGE}"
anything else should be after do_image_complete and be referencing DEPLOY_DIR_IMAGE.
Cheers,
Richard
Hi, Richard:
This is to handle the case IMAGE_CMD_yyy depending on IMAGE_CMD_xxx, while IMAGE_CMD_yyy need access the result generated by IMAGE_CMD_xxx in ${IMGDEPLOYDIR}, and IMAGE_CMD_yyy itself wanna generate a result in ${IMGDEPLOYDIR} as well, to be able to be covered by sstate.
In this case, IMAGE_CMD_yyy should not get the IMAGE_CMD_xxx's result from ${DEPLOY_DIR_IMAGE} because that might not be generated by the latest bitbake run.
And the result of IMAGE_CMD_xxx might not exit in ${IMGDEPLOYDIR} when rm_work is chosen in the following scenario:
1 bitbake image (IMAGE_CMD_xxx and IMAGE_CMD_yyy both run at the first time) 2 Change some code in IMAGE_CMD_yyy, but do not touch IMAGE_CMD_xxx 3 bitbake image
in the step 3, the result generated by IMAGE_CMD_xxx would not be in ${IMGDEPLOYDIR} if rm_work is chosen (Because it's removed in step 1), so this patch mainly aims to fix this problem.
//Ming Liu
Quoted 25 lines · On Fri, 2019-08-02 at 17:12 +0800, [email protected] wrote:
Quoted 4 lines
From: Ming Liu <[email protected]>
The results in ${IMGDEPLOYDIR} generated by a image task could be used by later image tasks.
No they shouldn't.
Quoted 4 lines
, hence they should be excluded by
rm_work. Or else when a image task gets rerun but one of its dependent image task does not, then the results generated by this dependent image task would not be there in ${IMGDEPLOYDIR}.Nothing should be poking into IMGDEPLOYDIR outside of the image tasks.
do_image_complete[sstate-inputdirs] = "${IMGDEPLOYDIR}" do_image_complete[sstate-outputdirs] = "${DEPLOY_DIR_IMAGE}"
anything else should be after
do_image_completeand be referencingDEPLOY_DIR_IMAGE.Cheers,
Richard