Re: "fatal: git-write-tree: error building trees" from `git stash`

[Jeff King <peff@peff.net>]

On Thu, Dec 27, 2012 at 01:55:56PM -0500, Alex Vandiver wrote:

> On Thu, 2012-12-27 at 10:51 -0800, Junio C Hamano wrote:
> > > $ git stash
> > > foo: needs merge
> > > foo: needs merge
> > > foo: unmerged (aeaa7e5e87cf309a7368d5d92a71c1f9e6a8c9e7)
> > > foo: unmerged (a77fa514de2720c72c1a861de098595959a2c97a)
> > > foo: unmerged (4a622d2b991f1a19ba7be313a46dc6f03692cd0a)
> > > fatal: git-write-tree: error building trees
> > > Cannot save the current index state
> > 
> > This is totally expected, isn't it?
> > 
> > You do not save state in the middle of a conflict with "git stash"
> > (instead, you would "git stash" away your own work in progress
> > before you start operation that may create and leave conflicts).
> 
> Apologies for not being clear.  While being unable to stash is not
> unexpected, perhaps, "Cannot stash while resolving conflicts" or similar
> would be more understandable to the end user than the above.

Yeah, I think the outcome is reasonable, but that message is just
horrible. Something like this might be better:

diff --git a/git-stash.sh b/git-stash.sh
index 688e259..7ea425c 100755
--- a/git-stash.sh
+++ b/git-stash.sh
@@ -217,6 +217,12 @@ save_stash () {
 
 	stash_msg="$*"
 
+	if ! git diff-index --cached --diff-filter=U --quiet HEAD; then
+		echo >&2 "fatal: unable to stash unmerged entries:"
+		git diff-index --cached --diff-filter=U --name-status HEAD
+		exit 1
+	fi
+
 	git update-index -q --refresh
 	if no_changes
 	then

but I suspect it is not sufficient:

  1. There are other code paths that will end up in write-tree which
     should probably be protected, too.

  2. Unmerged entries are only one reason that write-tree might fail.
     It's OK not to catch them all (since ultimately write-tree will
     complain if need be), but we may want to also handle intent-to-add
     entries with a nicer message.

-Peff