Macros don't behave as expected when macro register was written with a yank command vs macro recording command

[adlerd]

Issue type

• Bug report

Environment

Emacs version: GNU Emacs 25.3.1 (x86_64-pc-linux-gnu) of 2018-03-29
Operating System: Linux gentoo unstable (May 2018)
Evil version: 1.2.13
Evil installation type: MELPA (via portage)
Graphical/Terminal: terminal (via ssh, tmux, etc.)
Tested in a make emacs session (see CONTRIBUTING.md): No

Reproduction steps

• Start Emacs
• :new
• insert "ifoo^[", without quotes, with ^[ being the actual escape character (via C-v C-[)
• yank the line into a register, without the newline; i.e., 0"ay$
• execute this register as a macro (@A)

Expected behavior

"foo" should be inserted after point and we should return to Normal mode

Actual behavior

"foo" is inserted but we remain in Insert mode

Further notes

• The same repro steps in Vim 8.0 work as expected; probably in a lot of other vi's too
• I think evil-execute-macro should sanitize its input w/r/t escape just like evil-ex-normal does before passing to execute-kbd-macro; but I'm really not sure what the other consequences of this would be.
• Here's some alternative repro steps which make this look more like a register bug than a macro bug, in case that seems more problematic, since the problem is that paste+yank isn't idempotent:
  • record a macro into register a using the key sequence qaifooq
  • on a fresh line, paste register a then yank the line back into the same register ("ap0"ay$)
  • execute the macro
  • results: same as above. From a vim user's perspective, the register contents shouldn't have changed; from evil's perspective, they are subtly different.

[wasamasa]

The core issue here is that Evil reuses Emacs macros. Those can be represented as strings or vectors. Vectors cannot be pasted as is, so you'd need to translate them to strings, but then you'd also need to parse them which is anything but trivial. Vim goes for strings only and terminal-specific escapes, I don't really want to reimplement that, just to be able to yank keyboard macros into registers and re-execute them.

Anyway, my opinion on this is that when in Rome, do as the Romans do. For keyboard macros this means to use the dedicated editor and use elisp code to restore them programmatically. More thoughts on this can be found in #1038 and #1040.

[tttuuu888]

I found this same issue and my case was modification of existing macro and it worked as described here.
I also saw more weird behavior with more complex macro modification. For example, when I changed macro qi" <Esc> A " <Esc>q to qi" <Esc> A.h " <Esc>q (test to test.h) and yank to register, it workd pointlessly. (moved cursor several lines up)

When it comes to only modifying macro, We have good Emacs macro editor(kmacro-edit-macro). I think it will be useful If we can somehow use Emacs macro editor for modifying evil macro. Currently we can change last evil macro with kmacro-edit-macro but it is only effective when we execute macro with kmacro-call-macro.(of course modification works properly as well)