[Patch] mcedit overwrite mode buggy in UTF-8

[egmont]

Start mcedit 4.7.0.3 in a fully UTF-8 environment, turn on overwrite mode (F9 -> Edit -> Toggle ins/overw).

Type a couple of English letters, move the cursor back, and then type an accented letter.

Notice that typing the accented letter does not only overwrite the letter under the cursor, it also swallows the next character (or next two for 3-byte UTF-8 sequences, etc).

Interestingly though, in the mean time overwriting an accented letter with a plain English one works as expected, and overwriting an accented one with another accented one (of same byte length) is buggy.

(Seems to me that first as many _characters_ are removed as the number of _bytes_ in the letter to be inserted, and then the new letter is inserted correctly.)

Overwrite should always replace one letter with another letter, increasing/decreasing file size as necessary.

[egmont]

My assumption (in parentheses above) was right.

src/editor/edit.c has this function:

void
edit_execute_cmd (WEdit *edit, unsigned long command, int char_for_insertion)

and this function calls this in case of overwrite mode:

edit_delete (edit, 0);

whereas the boolean 0 means to remove a whole character, not just a byte (see the comment and implementation inside edit_delete).

The problem is: even though I'd expect char_for_insertion to hold the Unicode character to be inserted (based on the function's signature), this is not true. char_for_insertion is a single byte, and edit_execute_cmd is called individually for each byte of the UTF-8 sequence to be inserted.

I attach a quick and dirty patch which only calls edit_delete() on the first character of an UTF-8 sequence.

The real proper fix would probably require a major rewrite whereas the input handler composes full Unicode characters before actually modifying the file. Until this is done by someone, I believe my patch is perfectly reasonable, and should work perfectly if the input is really valid UTF-8 (which it should be).

[egmont]

a fix

[egmont]

Friendly ping...

A very simple patch is waiting to get applied - I believe it shouldn't take more than 5 minutes to verify it.

[angel_il]

Friendly pong...
1 packets transmitted, 1 received, 0% packet loss, time 4 days

you are right, patch is really small, but i dont sure that is good solution... i try search another way...

[angel_il]

need use edit->charpoint !

[angel_il]

branch: 2081_edit_overwrite_fix
changeset: a528eaa90f74fb618c97ebabaee7e566b03afb22

[egmont]

Oh yep, I didn't know about edit->charpoint but clearly this is the right way to go. Thanks!

[slavazanko]

Vote here.

Please note: apply changeset:b05422bf07e7a3e340f9d8a1c7cd88a53f8f4604 after rebase on master branch (to avoid conflicts) and before merge into 'master'.

[angel_il]

Fixed: f0b54977303b60f474b8ca84ab31b8f4a939ff92

[egmont]

Thanks guys!
Have you considered cherrypicking it to the 4.7.0/4.7.1 series? I think it would be nice to have the fix there too.

[angel_il]

Cherry-picked to 4.7.0-stable. 9b4714b2f419ca9c2c11053bcc28eb2a4fe71483