git rebase bug?

Hi,

I got a really weird result from a rebase today, and I'm wondering if that's a corner case or if that could be considered a bug in rebase.

This is reproducible with the following setup: $ git clone git://git.debian.org/pkg-mozilla/xulrunner/experimental xulrunner-1.9.2 $ cd xulrunner-1.9.2 $ git remote add upstream git://git.debian.org/pkg-mozilla/upstream $ git fetch upstream xulrunner/2.0:xulrunner/2.0 $ git checkout -b test upstream/1.9.2.4 $ git cherry-pick 67e469ef725ac3f4cdf043809066c353e6843db4

The patch that is cherry picked here has the following stats:
 security/manager/ssl/public/Makefile.in            |    1 +
 security/manager/ssl/public/nsIBadCertListener.idl |  155 ++++++++++++++++++++
 security/manager/ssl/src/nsNSSIOLayer.cpp          |  103 +++++++++++++-
 security/manager/ssl/src/nsNSSIOLayer.h            |    8 +

$ git rebase --onto xulrunner/2.0 upstream/1.9.2.4 test

The resulting commit has the following stats:
 security/manager/ssl/public/Makefile.in        |    1 +
 security/manager/ssl/src/nsNSSIOLayer.cpp      |  103 ++++++++++++++++-
 security/manager/ssl/src/nsNSSIOLayer.h        |    8 ++
 xulrunner/examples/simple/content/contents.rdf |  155 ++++++++++++++++++++++++

See how the security/manager/ssl/public/nsIBadCertListener.idl file that was created by the original patch is created as xulrunner/examples/simple/content/contents.rdf.

Please note that as the xulrunner/2.0 commit has no parent, I also tried grafting it on top of upstream/1.9.2.4, which didn't change anything.

So, corner case or definite bug?

Cheers,

Mike

PS: this all is with current master