Re: [BuildStream] Plans for workspaces and incremental builds

[William Salmon <will salmon codethink co uk>]

On 10/10/2019 10:16, Darius Makovsky via buildstream-list wrote:
> On 02/10/2019 13:43, William Salmon via buildstream-list wrote:
> > [...]
> >
> > Could we remove some of the friction hear? I would like to be able to
> > create a workspace from a failed build if I have cached buildtrees
> > enabled.
>
> I would expect that is possible but would require a bit of work. In theory I
> don't see why we can't do that.

We already had this for a while in `master` i hope we dont lose this.

> > Also can the final build with out a workspace realize that it
> > is the same as the already build workspace?
>
> That would be more difficult than the above. Again, *in theory* that is
> possible but essentially that would mean trying to save the workspace config
> for recovery in a non-workspaced state and then remapping a build tree to a
> new ref since the workspaced source ref is not going to be the same as the
> non-workspaced source ref.

Fiar.

> > [...]
> > When calculating the diffs don't forget that there might not be any
> > change to the source it might only be the bst file that changed. And
> > don't forget to handle soft reset properly/better.

Sorry for more clarity  `bst file that changed` I meant `the contents of 
the bst file that changed`

> If you have changed the bst then is it still the same workspace?

A work space is currently linked to a file name so if you change the 
contents of that file then yes you still have the same workspace.

> Also, I'm not even sure of the purpose/necessity of soft reset in the current
> implementation.

yes. whether or not you changed the contents of the bst file, some 
changes to the source on there own may need the configure steps to be re 
run so how ever you implement this workspace need to be soft reset-able.
Also the act of running .configure can be slow for big software so we 
dont want to lose this altogether.