[oe] Summary of an unwanted merge and what I did
Holger Freyther
zecke at selfish.org
Mon Apr 28 11:44:32 UTC 2008
Hey,
I love countries where they block the SMTP port and somehow even if I did sent
two mails on this.
So here my summary:
I had both Openembedded and Openmoko in the same database and innocently
typed mtn merge, got annoyed with the ncc on gsmhandset.state files,
completed the merge, pushed. I only had a look at the last rev of the two
heads and it was a DEPENDS fixlet to neod which looked about right, so
decided to do the merge. Only afterwards I saw I just merged the qtopia-phone
crap.
So what did I learn about mtn:
- You can not disapprove a merge!
- So I decided to suspend the merge, the ncc preparation and the head of OM.
This lead to a tree without any heads. :)
- Then I used the last not OM head I had (one ancestor of the merge) and did
a change and created a new head.
I assumed this situation was fixed, we have OM history in the OE.db which is
not bad, but still the Qtopia phone recipes are not within OE. Everything
looked fine until I got poked on irc.
It looks like someone is using a mtn without the support for the suspend cert,
and he did the merge again.
I could suspend the merge, create a new head without the merge but then
someone without a mtn with the suspend cert support will do a merge... and I
could suspend the merge, create a new head wihtout the merge but then someone
without a mtn with the suspend will do a merge... I love infinite recursion.
I have no idea how to fix this situation without ditching monotone. E.g. If
mtn would transfer the client version number and we could reject this with a
lua hook we would be set, but I don't remember seing such thing.
ideas? I'm tempted to say sorry but I think I'm not the one to blame (a tool
is) :)
z.
More information about the Openembedded-devel
mailing list