[yocto] sstate-cache question

Gary Thomas gary at mlbassoc.com
Wed May 31 20:44:49 PDT 2017


On 2017-06-01 05:28, Paul Eggleton wrote:
> Hi Gary,
> 
> My apologies, I just realised I never sent this reply.
> 
> On Tuesday, 23 May 2017 5:27:57 PM NZST you wrote:
>> On 2017-05-22 22:35, Paul Eggleton wrote:
>>> On Tuesday, 23 May 2017 2:53:45 AM NZST Gary Thomas wrote:
>>>> I have a build where I've never manually removed anything from
>>>> the sstate-cache and this same build has been used for hundreds
>>>> of builds over the last 18 months.  I just tried to find out why
>>>> gcc-cross-arm had to be rebuilt (it seems to happen almost every
>>>> time I update my Poky/Yocto tree (master)).  Here's what I got:
>>>>
>>>> $ bitbake-diffsigs -t gcc-cross-arm compile
>>>> Hash for dependent task gcc/gcc-cross_6.3.bb.do_configure changed from
>>> 208373dd9ae01101a26a9412eb50b110 to
>>>> d65095d4b9aff89f6990bd17c0ab210b
>>>> Unable to find matching sigdata for /local/poky-cutting-edge/meta/
>>>> recipes-devtools/gcc/gcc-cross_6.3.bb.do_configure
>>>> with hashes 208373dd9ae01101a26a9412eb50b110 or
>>> d65095d4b9aff89f6990bd17c0ab210b
>>>>
>>>> So if I didn't remove those files, where did they go?  Am I doing
>>>> something wrong running this tool?  (running the same command for
>>>> qemu-native seemed to work correctly)
>>>
>>> Is this with master, pyro, morty or some other version?
>>
>> Poky master: 2568e18701fb20d7105eb6e4929f3aff4b9f9c06
> 
> OK so this is definitely after my recent fixes then.
>   
>>> If you search for files with those hashes in their name do they show up?
>>
>> Only .siginfo files:
>> $ find sstate-cache/ -name "*d65095d4b9aff89f6990bd17c0ab210b*"
>> sstate-cache/universal/d6/sstate:gcc-cross-arm:x86_64-amltd-linux-gnueabi:
>> 6.3.0:r0:x86_64_arm:3:d65095d4b9aff89f6990bd17c0ab210b_configure.tgz.siginfo
>> $ find sstate-cache -name "*208373dd9ae01101a26a9412eb50b110*"
>> sstate-cache/universal/20/sstate:gcc-cross-arm:x86_64-amltd-linux-gnueabi:
>> 6.3.0:r0:x86_64_arm:3:208373dd9ae01101a26a9412eb50b110_configure.tgz.siginfo
>>
>> However, I do find some in tmp/stamps:
>> $ ls tmp/stamps/x86_64-linux/gcc-cross-arm
>> 6.3.0-r0.do_build.41aca0d85971087f4675d2f504642ce4
>> 6.3.0-r0.do_compile.b0eae7e91833efd5e7eceaedbc59983e
>> 6.3.0-r0.do_compile.sigdata.9a06aa40225be30babaceb7673cef0a6
>> 6.3.0-r0.do_compile.sigdata.b0eae7e91833efd5e7eceaedbc59983e
>> 6.3.0-r0.do_configure.d65095d4b9aff89f6990bd17c0ab210b
>> 6.3.0-r0.do_configure.sigdata.208373dd9ae01101a26a9412eb50b110
>> 6.3.0-r0.do_configure.sigdata.d65095d4b9aff89f6990bd17c0ab210b
>> 6.3.0-r0.do_fetch.1dd3c87b3e509c20fa4ffe61d6dc3b41
>> 6.3.0-r0.do_gcc_stash_builddir.063822f91220bb40cef0696eb604a568
>> 6.3.0-r0.do_gcc_stash_builddir.sigdata.063822f91220bb40cef0696eb604a568
>> 6.3.0-r0.do_gcc_stash_builddir.sigdata.28144d3c18c0a09f23f3b0c5e80f049d
>> 6.3.0-r0.do_install.0f96c5448dc9360ca3a24d551ad3da89
>> 6.3.0-r0.do_install.sigdata.0f96c5448dc9360ca3a24d551ad3da89
>> 6.3.0-r0.do_install.sigdata.f2e4d12289c5039854081a832dc8bc0e
>> 6.3.0-r0.do_populate_lic_setscene.d2c9226ddb0de5277e2347b69ef84664
>> 6.3.0-r0.do_populate_sysroot.15afc3bb019a6c0c1285cbda46cb32b0
>> 6.3.0-r0.do_populate_sysroot.sigdata.15afc3bb019a6c0c1285cbda46cb32b0
>> 6.3.0-r0.do_populate_sysroot.sigdata.a0c5b715ae9217f02318beff6987d169
>> 6.3.0-r0.do_prepare_recipe_sysroot.150112323551011e0e87f99f47ad79c4
>> 6.3.0-r0.do_prepare_recipe_sysroot.sigdata.150112323551011e0e87f99f47ad79c4
>> 6.3.0-r0.do_prepare_recipe_sysroot.sigdata.4a528becf11832cc3b8f6fa479e98210
>>
>> I guess the sstate-cache info alone is not sufficient to use 'bitbake
>> diffsigs'?
> 
> It ought to be. At face value it should be finding the files in either place
> - we'd need to debug the code to find out why it isn't. Is that something you
> might be able to help me with?
> 

Happy to give it a go, just let me know what you need

Thanks

-- 
------------------------------------------------------------
Gary Thomas                 |  Consulting for the
MLB Associates              |    Embedded world
------------------------------------------------------------



More information about the yocto mailing list