Richard Hipp wrote:
On 5/7/16, Joel Bruick<j...@joelface.com> wrote:
The "merge-renames" branch is my effort to make any and all rename
scenarios mergeable.
I've been testing the merge-renames branch by redoing all historical
merges in Fossil itself, using both trunk and merge-renames, and
checking that they get the same result.
The new merge-renames Fossil does not seem to be doing the
https://www.fossil-scm.org/fossil/info/346d62a4118d868b merge
correctly. The problem occurs on the setup/fossil.nsi file. It
appears that trunk Fossil correctly incorporates the one-character
change to setup/fossil.nsi that occurred on the cleanX. However, when
I repeat the merge using merge-renames Fossil, it is not pulling in
that change and thus the edit is lost.
Thanks for catching this. The problem here was that I wasn't resetting
the pivot-finding algorithm before finding nid, which in this case
caused pivot_find() to return a version that was a first-parent ancestor
of V, but not of M or P. This causes problems when we search for renamed
files, as find_filename_changes() calls path_shortest() with directOnly
set to true, so it could only find renames for N->V.
Let me know if you find any other problems. I still need to investigate
one or two other things that I have a hunch might not be handled
correctly but don't have a test for yet.
_______________________________________________
fossil-dev mailing list
fossil-dev@mailinglists.sqlite.org
http://mailinglists.sqlite.org/cgi-bin/mailman/listinfo/fossil-dev
