Pulling a 'thin strand of development' in the presence of subrepos

[Alistair Bell]

Hi all,

We have a situation where we have 'clean' and 'dirty' repos. The 'dirty' repos are supersets of the 'clean' repos. We're also using subrepos.

When a user wants to make his work part of the 'clean' repo, he pushes it to the 'dirty' repo (sometimes creating a new head) and passes the changeset ID to a regression engine, which pulls just that changeset ID and its ancestors from the 'dirty' repo using 'hg pull -r <ID> dirty -u'.

HOWEVER, although the pull only pulls the ID and its ancestors from the main repo, the update pulls the entirety of the subrepo of dirty (not just the ancestors of the changeset I need). Is this by design or by oversight? 

Is there a better way to do what I'm attempting?

It looks to me like code could be added in postincoming to change this, but I don't particularly want to start messing with stuff if there's a better way.

Any ideas?