master eSDK hash problem with METADATA_REVISION and base-files:do_install
Building an eSDK (all stock config):
MACHINE=qemuarm bitbake core-image-minimal -c populate_sdk_ext
I then install it, and get:
Checking sstate mirror object availability: 100% |#####################################################################################################################|
Time: 0:00:00
WARNING: The base-files:do_install sig is computed to be587ccfa9299e26094ae342f1ce3d782bfaf776c639013428a511770c3828d49a, but the sig is
locked to 0c35f266f1e79b1f34260df1b8d677f7eb74a7f7241155601ed73a014c8e286c in
SIGGEN_LOCKEDSIGS_t-qemuarm
Loading cache: 100% |##################################################################################################################################################|
Time: 0:00:00
Initialising tasks: 100% |#############################################################################################################################################|
Time: 0:00:00
Running a bitbake diffsigs on this file, I get:
basehash changed from cffec52acb67f62fe3812d0d42d52b1081f9150386f43f0426f5469b4a1a6a25 to 332259212c5e9a4ed2a7fe25b8ea77cb54fb6b82f7deed90179e145f1d0cf1b6 Variable METADATA_REVISION value changed from '3210cabe56d6cf6d51638e6f31f93110e40ab2cd' to '<unknown>'
So it appears that some of the recent work on METADATA_REVISION has broken hash computations in the base-files package due to the value being 'unknown' in one, and set in the other.
--Mark
Hi Mark,
I think this has been fixed by the following commit.
"""
commit ff2ad51b801fd62e2abbc573ba2c9ee8fdc7e012
Author: Chen Qi <[email protected]> Date: Fri Mar 5 18:10:40 2021 +0800
populate_sdk_ext: record METADATA_REVISION
As we delete the .git/ directory, it's impossible to getMETADATA_REVISION
inside eSDK. Because of this, we meet the following warning when
installing eSDK.
WARNING: The base-files:do_install sig is computed to be16b9d96148d45de183cc94667aae016ec7d102d48255456381e718cd4bbd0aa0, \
but the sig is locked to6eb0dcaed504282becee94662481d79264db920dee1f7deda18230133fff8f36 in SIGGEN_LOCKEDSIGS_t-qemux86-64
So we record METADATA_REVISION in eSDK generation time to fix this
problem.
"""
Best Regards,
Chen Qi
Quoted 40 lines · Mark Hatle
Building an eSDK (all stock config):
MACHINE=qemuarm bitbake core-image-minimal -c
populate_sdk_extI then install it, and get:
Checking sstate mirror object availability: 100% |#####################################################################################################################|
Time: 0:00:00
WARNING: The base-files:do_install sig is computed to be587ccfa9299e26094ae342f1ce3d782bfaf776c639013428a511770c3828d49a, but the sig is
locked to0c35f266f1e79b1f34260df1b8d677f7eb74a7f7241155601ed73a014c8e286cin
SIGGEN_LOCKEDSIGS_t-qemuarm
Loading cache: 100% |##################################################################################################################################################|
Time: 0:00:00
Initialising tasks: 100% |#############################################################################################################################################|
Time: 0:00:00Running a bitbake diffsigs on this file, I get:
basehash changed from
cffec52acb67f62fe3812d0d42d52b1081f9150386f43f0426f5469b4a1a6a25to332259212c5e9a4ed2a7fe25b8ea77cb54fb6b82f7deed90179e145f1d0cf1b6VariableMETADATA_REVISIONvalue changed from '3210cabe56d6cf6d51638e6f31f93110e40ab2cd' to '<unknown>'So it appears that some of the recent work on
METADATA_REVISIONhas broken hash computations in the base-files package due to the value being 'unknown' in one, and set in the other.--Mark
I was using master-next (version was right before I sent the email), and it still failed.
Looking at things, there doesn't appear to be any METADATA_REVISION set in the normal build, but there is one set inside of the eSDK. So it looks like some corner case may have been missing.
If you notice it says <unknown>. Diffsigs I had, I don't know how the 'default' version should be generated or not -- but it was definitely different inside and outside the eSDK.
--Mark
Quoted 74 lines · Chen Qi
Hi Mark,
I think this has been fixed by the following commit.
"""
commitff2ad51b801fd62e2abbc573ba2c9ee8fdc7e012
Author: Chen Qi <[email protected]>
Date: Fri Mar 5 18:10:40 2021 +0800
populate_sdk_ext: recordMETADATA_REVISION
As we delete the .git/ directory, it's impossible to getMETADATA_REVISIONinside eSDK. Because of this, we meet the following warning when installing eSDK.
WARNING: The base-files:do_install sig is computed to be16b9d96148d45de183cc94667aae016ec7d102d48255456381e718cd4bbd0aa0, \
but the sig is locked to6eb0dcaed504282becee94662481d79264db920dee1f7deda18230133fff8f36in SIGGEN_LOCKEDSIGS_t-qemux86-64
So we recordMETADATA_REVISIONin eSDK generation time to fix this problem.Signed-off-by Chen Qi <[email protected]>Signed-off-by Richard Purdie <[email protected]>"""
Best Regards,
Chen QiQuoted 38 lines · Mark Hatle
Building an eSDK (all stock config):
MACHINE=qemuarm bitbake core-image-minimal -c
populate_sdk_extI then install it, and get:
Checking sstate mirror object availability: 100% |#####################################################################################################################|
Time: 0:00:00
WARNING: The base-files:do_install sig is computed to be587ccfa9299e26094ae342f1ce3d782bfaf776c639013428a511770c3828d49a, but the sig is
locked to0c35f266f1e79b1f34260df1b8d677f7eb74a7f7241155601ed73a014c8e286cin
SIGGEN_LOCKEDSIGS_t-qemuarm
Loading cache: 100% |##################################################################################################################################################|
Time: 0:00:00
Initialising tasks: 100% |#############################################################################################################################################|
Time: 0:00:00Running a bitbake diffsigs on this file, I get:
basehash changed from
cffec52acb67f62fe3812d0d42d52b1081f9150386f43f0426f5469b4a1a6a25to332259212c5e9a4ed2a7fe25b8ea77cb54fb6b82f7deed90179e145f1d0cf1b6VariableMETADATA_REVISIONvalue changed from '3210cabe56d6cf6d51638e6f31f93110e40ab2cd' to '<unknown>'So it appears that some of the recent work on
METADATA_REVISIONhas broken hash computations in the base-files package due to the value being 'unknown' in one, and set in the other.--Mark