[OE-core] overuse of bitbake flags considered harmful

[Richard Purdie]

On Thu, 2012-09-27 at 13:22 +0100, Phil Blundell wrote:
> I was mildly amused to discover, the other day, that the output of
> "bitbake -e rpm | grep PACKAGECONFIG" is not quite as one might naïvely
> expect.  
> 
> I guess the right response to this is perhaps that bitbake ought
> nowadays to find some other way to store its own internal state, since
> the idiom of using flags is pretty well established in oe-core.

For those who don't realise what's going on here, the output is:

python PACKAGECONFIG () {
db bzip2 zlib beecrypt openssl libelf python}

since "python" is a flag bitbake uses to identify python functions too.

I suspect the better solution would be to use "_python" internally since
we already have a convention within bitbake of prefixing internally used
flags with "_".

http://git.yoctoproject.org/cgit.cgi/poky/commit/bitbake/lib/bb/data_smart.py?id=49bc007c80e429090c4db51e30c5ce27e8a72558
http://git.yoctoproject.org/cgit.cgi/poky/commit/bitbake/lib/bb/data_smart.py?id=980d658f3f1e6c5b459f741385533a11d35009ca

This would mean the metadata shouldn't use that namespace but I think we
can live with that.

Cheers,

Richard