SRCREV and git packages

[Werner Almesberger]

A few more thoughts:

I wrote:
> Q: As above, but we want any version for which A <= sequence N <= B.
> 
>   Assuming that version A exists at the time the package description is
>   made, the package description maker could include a hint in the
>   description that informs what hash corresponds to sequence A.

This may actually not be a good example. If we want to talk about
ranges of versions, we might as well use the hash instead of the
sequential number, so no hint is needed.

However, in real life, one would probably want to refer to a symbolic
name, such as "version 2.1". Mapping this to a revision in the SCM is
a different kind of problem, which isn't unique to git.

> Q: As above, but we want the _latest_ verison which A <= sequence N <= B.
> 
>   I think this is the point where we need a full dictionary.

That is, if we _predict_ B. Given that SCM revisions number are in
general not very predictable, this may not be a relevant scenario
either.

If version B exists at the time the package is made, this can be
treated just like the previous example.

To take this one step further, if I got my examples right and there
aren't any vastly different queries I've overlooked, it seems that
everything can be quite adequately expressed just with hashes
inside the bitbake world.

I realize that recording the additional information needed to handle
hashes efficiently, and expressing the queries in the appropriate way
may not be trivial. But I think enabling OE to treat git repositories
as "natively" as possible is probably something you'd want to consider
anyway.

You'd still need sequential numbers when making packages, so that
opkg or similar can tell what's old and what's new, but since you
have access to the repository at that time, that's easy to do.


Going back to the current scenario with sequential numbers generated
from git-rev-list, there's one issue that was raised that isn't clear
to me: why would one have to download repositories for packages that
are not used at all ?

Would this just be a side-effect of how generation of the sequential
numbers is implemented, or are there real-life cases where one has
to know the sequential number of a package that's of no further
interest ?

- Werner