ed: quit: warn buffer modified after failed e cmd v1 REJECTED

Curtis Arthaud: 1
 quit: warn buffer modified after failed e cmd

 1 files changed, 0 insertions(+), 1 deletions(-)

[PATCH ed] quit: warn buffer modified after failed e cmd Export this patch

Before this patch we had:

*a
hello
.
*e another_file
Warning: Buffer modified
*q # happily quits, losing "hello"

Signed-off-by: Curtis Arthaud <uku82@gmx.fr>
---
 file.ha | 1 -
 1 file changed, 1 deletion(-)

diff --git a/file.ha b/file.ha
index c3068e7..6396bbd 100644
--- a/file.ha
+++ b/file.ha
@@ -9,7 +9,6 @@ fn edit(s: *Session, cmd: *Command, forced: bool) (void | Error) = {
 	assert_noaddrs(s, cmd.linenums)?;

 	if (!forced && s.buf.modified && !s.warned) {
-		s.warned = true;
 		return WarnBufferModified;
 	};

--
2.44.0

Byron Torres <b@torresjrjr.com>

6 months ago

There's a debate here whether the current behaviour is correct. Here's
the relevant excerpt from ed(1p):

Byron Torres <b@torresjrjr.com>

6 months ago

It is not super clear whether ['e', 'q'] or ['q', 'e'] counts as
"repeating". I think it is inferred from the first paragraph that it is
the destroying of the editor buffer that is more important than the
command that causes that destruction. GNU ed follows this
interpretation, and I agree with it.

	$ ed --version | head -n1
	GNU ed 1.20.1
	$ ed preexistingfile
	H
	P
	*a
	hello
	.
	*e
	Warning: buffer modified
	*q
	$ ed preexistingfile
	H
	P
	*a
	hello
	.
	*q
	Warning: buffer modified
	*e
	419
	*q
	$ 

Besides, removing that line of code I think isn't correct. You'd have to
implement `Session.warned_edit` and `Session.warned_quit`.