Comment # 3 on bug 954556 from Tony Mechelynck

Had it again, but my workflow is improving a little:

1) "hg fetch --switch-parent" still does a pull but aborts the merge:

$ hg fetch --switch-parent || echo 'exit status' $? ; date
pulling from https://bitbucket.org/vim-mirror/vim
searching for changes
sampling from both directions
adding changesets
adding manifests
adding file changes
added 13 changesets with 34 changes to 23 files (+1 heads)
resolving manifests
0 files updated, 0 files merged, 0 files removed, 0 files unresolved
merging with 7275:8896150aba23
resolving manifests
runtime/doc/tags: untracked file differs
abort: untracked files in working directory differ from files in requested
revision
exit status 255
Thu 19 Nov 22:32:03 CET 2015

2) then after updating to the latest common ancestor and back to the newest
pulled changeset:

$ hg merge -r 7262
resolving manifests
local changed runtime/doc/tags which remote deleted
use (c)hanged version or (d)elete? d
removing runtime/doc/tags
[...]
103 files updated, 0 files merged, 1 files removed, 0 files unresolved
(branch merge, don't forget to commit)
$ hg commit -m 'merge local with https://bitbucket.org/vim-mirror/vim'
committing files:
[...]
committing manifest
committing changelog
committed changeset 7276:01ce4237faad

N.B.
- revision 7262 (decimal) was the tip before the fetch
- Under 13.2 the fetch succeeded, telling me "remote changed runtime/doc/tags
which local deleted" instead (since we were on the other branch)
- I'd prefer to see fewer files "updated" by the merge step, as it used to be
under 13.2. I think it makes a difference to the following "make" (does it
react the same way to a file left in place and to a file removed then added
back?).
- Strangely enough, the "hg log" entry for the merge commit mentions only one
file (the one I removed by answering d at the prompt), not 103.
- By doing "hg diff" between the new tip (the merge commit) and one of its
parents, I notice that some files listed in .hgignore (linkable objects from a
previous "make", apparently) have been "added". Is this normal? (The .hgignore
uses glob syntax, its first line is "syntax: glob", and no other line starts
with "syntax".)

It seems the merge succeeds in one direction and fails in the other one, while
obtaining (AFAICT) the same result in the working directory.