[oe] bitbake dependency cache problem

[Steffen Sledz]

On 05.03.2012 14:28, Richard Purdie wrote:
> On Thu, 2012-03-01 at 14:46 +0100, Steffen Sledz wrote:
>> I'm working with oe-classic and BitBake Build Tool Core version
>> 1.12.0, bitbake version 1.12.0.
>>
>> Because of some special development requirements we like to generate
>> the Package Version from the SVN Revision (not the Last Changed Rev!)
>> of the bitbake recipe in the local workspace.
>>
>> Therefor the recipe contains this:
>>
>> ---------------->snip<-----------------
>> PR = "r8"
>> PV = "svnr${@svn_revision(d)}"
>>
>> inherit svn-helper
>> ---------------->snip<-----------------
>>
>> The svn-helper.bbclass contains a little helper function to determine
>> the needed value:
>>
>> ---------------->snip<-----------------
>> def svn_revision(d):
>>      import subprocess
>>      bbpath = os.path.dirname(bb.data.getVar('FILE',d,1))
>>      return subprocess.check_output(["svn", "info", bbpath]).partition("Revision: ")[2].splitlines()[0]
>> ---------------->snip<-----------------
>>
>> After this changes i make a first build and everything works fine.
>> Assuming the SVN Revsion is 42 a package called foo-svnr42-r8.ipk is
>> generated.
>>
>> Now i make an "svn update" inside the workspace and the SVN Revision
>> increases to 66.
>>
>> A call of "bitbake foo" now results in an "Tasks Summary: Attempted
>> 1182 tasks of which 1182 didn't need to be rerun and 0 failed." an no
>> new ipg is generated. :(
>>
>> The log generated by "bitbake -DDDDDvvvvv foo" contains
>>
>> ---------------->snip<-----------------
>> DEBUG: providers for foo are: ['foo']
>> NOTE: checking PREFERRED_PROVIDER_foo
>> NOTE: checking PREFERRED_PROVIDER_foo-svnr42
>> NOTE: checking PREFERRED_PROVIDER_foo-svnr42-r8
>> ---------------->snip<-----------------
>>
>> Why does bitbake not respects the new PV and generates a
>> foo-svnr66-r8.ipk here???
> 
> Bitbake has no idea that the PV has changed and that it needs to ignore
> the cache value and reparse the metadata for that file. The similar
> example is when autorev is used with a source control system. You can
> can see the code for this in lib/bb/fetch2/__init__.py in get_autorev().
> In summary, if you set __BB_DONT_CACHE = "1" in your recipe, I think it
> will do what you expect, at the cost that it will reparse the recipe
> every time (which it has to in case PV changes).

This solution seems to work. :)

Thx,
Steffen

-- 
DResearch Fahrzeugelektronik GmbH
Otto-Schmirgal-Str. 3, 10319 Berlin, Germany
Tel: +49 30 515932-237 mailto:sledz at dresearch-fe.de
Fax: +49 30 515932-299
Geschäftsführer: Dr. Michael Weber, Werner Mögle;
Amtsgericht Berlin Charlottenburg; HRB 130120 B;
Ust.-IDNr. DE273952058