On 06/13/2016 05:41 AM, Stephen Finucane wrote:
Users can send a new revision of a single patch from a series. When
this happens there is no immediate context: one must grok the
previous series to understand that this new patch is in fact a minor
change to an existing series and not a new, standalone hanges. This is
particularly impactful for things like CI which _need_ to understand
this dependency trail in order to successfully apply and test a patch.
Resolve this issue through the "auto-completion" of series. This is
achieved by examining previous series and taking only the patches that
have not be replaced in the current version. For example, if a user
sends a three patch series and then submits a revision to patch two,
only patches one and three of the prior series will be used and the
resulting group presented as an entirely new series.
This is done entirely dynamically. The reason for this is that series
relationships themselves are dynamic: sometimes it may not be possible
to identify a relationship until the last patch of a new series is
submitted (say there's a three patch series and the first two parsed
patches are significantly reworked).
This won't catch situations where the order of a patch has been moved
or the patch has been significantly reworked. However, one can hope
users are not silly enough to do something like that.
Signed-off-by: Stephen Finucane <[email protected]>
I'm not sure I'm understanding this properly. Here's how I tried to test
this:
1) import this entire series with parsearchive.py
2) took this patch, and updated its headers to:
- have a new "Message-Id"
- make In-Reply-To and References point to the original 10/10 patch.
3) imported this new hand-crafted patch
The result was the new patch being added to the same series revision.
The way I read this commit message, I was expecting it to generate a new
SeriesRevision. Am I missing something?
_______________________________________________
Patchwork mailing list
[email protected]
https://lists.ozlabs.org/listinfo/patchwork
