Re: sscache signature?

[Richard Purdie <richard.purdie@linuxfoundation.org>]

On Wed, 2011-08-10 at 12:51 -0500, Kumar Gala wrote:
> the yocto wiki references mentions a bitbake-diffsigs
> 
> https://wiki.yoctoproject.org/wiki/Enable_sstate_cache
> 
> but not seeing that.  How can one compare signature or try and figure
> out why a cache item isn't being reused between cfgs.

What I've been referring to is a process something along the lines of:

$ bitbake bash -e | grep BASEHASH | sort > hashbefore
$ vi ../scripts/oe-buildenv-internal 
[Add SOCKS5_USER and SOCKS5_DOMAIN to BB_ENV_EXTRAWHITE]
$ export SOCKS5_USER=bar
$ cd ..; source oe-init-build-env b2
$ bitbake bash -e | grep BASEHASH | sort > hashafter
$ diff -u hashbefore hashafter
[no differences]
$ vi conf/local.conf
[ Add "export SOCKS5_USER" to file ]
$ bitbake bash -e | grep BASEHASH | sort > hashafter2
$ diff -u hashbefore hashafter2
[ Find hashes now changed (which is bad and what we need to avoid ]
$ vi ./meta/conf/bitbake.conf
[ Add SOCKS5_USER to BB_HASHBASE_WHIETLIST ]
$ bitbake bash -e | grep BASEHASH | sort > hashafter3
$ diff -u hashbefore hashafter2
[no differences again - good]

which proves the addition of the variable didn't result in changes
sstate checksums. The bitbake-diffsigs command can be used to find out
what caused changes but in this case we have a pretty good idea so I
didn't use it.

As mentioned on irc, adding it to BB_ENV_EXTRAWHITE isn't an issue but
since its not exported into any of bitbake's execution environment, I'm
at a loss to understand how this helps you. If you do also export the
value to include it in bitbake's execution environment it does change
the hash which is what I worried about and it needs excluding as
documented above.

Some further investigation is needed to figure out how BB_ENV_EXTRAWHITE
is enough as it shouldn't be.

Cheers,

Richard