Re: Manifests and metadata

[Tristan Van Berkom <tristan vanberkom codethink co uk>]

On Tue, 2017-11-07 at 10:51 +0000, Sam Thursfield wrote:
> On 07/11/17 06:15, Tristan Van Berkom wrote:
[...]
> I think this is a small enough issue that we can ignore it, but we need 
> to be careful about putting /too/ much into logs.

Agree, but I feel much more strongly about not putting too much into
artifact metadata.

To the point where perhaps the original manifest-in-metadata
optimization I posted on the issues is wrong (maybe instead BuildStream
should load the manifest as python objects by inferring it from the
extracted artifact only once, and avoid walking the filesys
afterwards).

Even if we need the attributes serialized in a checked out artifact in
order to reliably do GPG signing/verifying, that could be a flat list
of attributes which could be associated with an inferred, sorted walk
of the files/ checkout, without adding too much metadata.

> We also still need to think about a way of making access to the log 
> files convenient; currently I have do this by poking around in the 
> artifact cache directly. but as long as we document the procedure then 
> it should be OK for now.

Yeah, I've been wanting similar for a while, maybe we need to add
something like a new format to `bst show` to view a log.

Right now the closest we have is the interactive "log" choice when a
build fails (which I use and enjoy a lot).

Cheers,
    -Tristan