On 5/8/20 4:47 PM, Phillip Wood wrote:
Hi Konstantin
On 08/05/2020 11:35, Konstantin Kharlamov wrote:
As description says. To check that the 3-way merge should actually work subsitute `git apply -3` with `git am -3` in the following steps. This way it works.
# Steps to reproduce
$ git init
Initialized empty Git repository in /tmp/foo/.git/
$ echo hello > file1 && git add file1 && git commit -m "initial commit"
[master (root-commit) 8334093] initial commit
1 file changed, 1 insertion(+)
create mode 100644 file1
$ git checkout -b mybranch
Switched to a new branch 'mybranch'
$ echo bye > file1 && git add file1 && git commit -m "change file1 text"
[mybranch 1807900] change file1 text
1 file changed, 1 insertion(+), 1 deletion(-)
$ git format-patch -1 --stdout > 1.patch
$ git checkout master
Switched to branch 'master'
$ mv file1 file2
$ git add -u && git add file2 && git commit -m "renamed file1 to file2"
$ git apply -3 1.patch
error: file1: does not exist in index
Judging from the error message it never gets as far as trying the 3 way merge because it first tries to apply the patch which modifies file1 but file1 does not exist in the index. If it were to try a 3 way merge in this case then on one side of the merge you've modified file1 and on the other side it has been deleted so there would be a merge conflict. The reason 'git am' succeeds is that the rename detection kicks in and so it applies the patch to file2 rather than file1
Best Wishes
Phillip
Thank you. I wonder if it would make sense to `git aplly -3 1.patch` to behave similarly to `git am -3 1.patch && git reset HEAD^` (except without modifying reflog). I.e. so all heuristics `git am` makes use of would also apply to `git apply`
