While experimenting with using The Bot for TSC-99 (topic:inding-tlc-rpc), it produced this:

mdsal:55/69355/34,mdsal:62/69362/31,...

which is NOK, because the the multipatch job does not support this - it will cherry-pick both changes SEPARATELY into a fresh git clone; which is not what we want , and will fail as the 2nd is dependant on the 1st so cherry-pick fails.

I can think of 2 ways to make better tooling for this situation - let's discuss here and agree:

1. Either, I make the The Bot smart enough to detect parental ancestry of changes for the same project on a topic (that's probably doable), so that it spits out this PATCHES_TO_BUILD magic string instead of above:

mdsal=62/69362/31,...

This should work, with the minor disadvantage that if such patch series are not rebased, it could build outdated code (which the cherry pick avoids) and potentially give false results.

That could be addressed if the multipatch job would just do a rebase after checkout ...

2. the multipatch job, or it's successor in RELENG-102, could allow specifying a list of patches to cherry pick as above, and pick EACH of them into the SAME git clone. This would probably mean that The Bot will have to order them by their Parent, as the job probably can't (easily) do that?

skitt ecelgp zxiiro jluhrsen dfarrell07 any thoughts which way to go?