Code fix for the "slow ESC" key input in old variants?

[smbhax]

Old variants such as Hellband and Zangband process ESC key input only after a momentary delay, at least in Curses mode; the "tilde" key just below ESC does the same thing, but operates immediately.

Is there a general code fix for the slow ESC key input in these variants?

[Gwarl]

seems like it was fixed some time between vanilla 3.2.0 and vanilla 3.4.1. I daren't check 3.3.0 right now

[wobbly]

If its the same problem oangband has I get round it by macro-ing the escape key to the escape key. Don't know why that works, just seems to.

[smbhax]

seems like it was fixed some time between vanilla 3.2.0 and vanilla 3.4.1. I daren't check 3.3.0 right now

Intriguing!

If its the same problem oangband has I get round it by macro-ing the escape key to the escape key. Don't know why that works, just seems to.

Oh that's interesting. I tried doing some kind of ESC/tilde keymap in Zangband but couldn't seem to get it to take, same with an ESC-ESC one. I don't know my way around macros very well, though.

[smbhax]

Current Angband has a section in src/main-gcu.c starting with this disclaimer

	/* This might be a bad idea, but...
	 *
	 * Here we try to second-guess ncurses. In some cases, keypad() mode will
	 * fail to translate multi-byte escape sequences into things like number-
	 * pad actions, function keys, etc. So we can hardcode a small list of some
	 * of the most common sequences here, just in case.

and including

		case 27:        i = ESCAPE; break;

and similar statements for arrow keys, keypad keys, etc. Zangband's src/main-gcu.c has no analogous section as far as I can tell. Perhaps this section has something to do with it?

Angband's also has

	/* We do it like this to prevent a link error with curseses that
	 * lack ESCDELAY. */
	if (!getenv("ESCDELAY")) {
#if _POSIX_C_SOURCE < 200112L
		static char escdelbuf[80] = "ESCDELAY=20";
		putenv(escdelbuf);
#else
		setenv("ESCDELAY", "20", 1);
#endif
	}

which looked tantalizing, but just dropping it in Zangband did not magically cure the sluggish ESC key. ;; )

[smbhax]

seems like it was fixed some time between vanilla 3.2.0 and vanilla 3.4.1. I daren't check 3.3.0 right now

I went to try out the releases in this range but for all of them got "multiple definitions" compile errors concerning z-file.h or z-term.h, so I have been of no use.

[smbhax]

In build notes, the only thing that jumped out at me--not that I know that I'm looking for--is that 3.4.0 included "Upgrade to UTF-8."

Oh and "Fix GCU port to recognise backspace and del as distinct keys."

[smbhax]

Modern ESC key support was added to Angband by stefanor@cox.net on January 22, 2009 in https://github.com/angband/angband/c...f76cc30dcb9bc5

Their description of that commit is

* Make pseudo graphics the default for main-gcu, as they are for the other ports.
* Changed the default macros to use the new pre-processed keys, thus mostly eliminating the "escape delay" and being more portable.
* Implement support for the curses "alternate character set", used for pseudographics e.g. by nethack 'DECgraphics'. The high bit of characters is repurposed for this, hopefully the Eastern Europe community won't mind.
* We now use curses to parse keys, which does a much better job than the old "abuse of macro system". More portable and significantly faster, especially for Escape.
* Unprintable characters are blotted with ?s, removing another irritating display glitch from the visual editor.

It DOES include that "ESCDELAY" section in main-gcu.c.

It went out in Angband 3.1.1, 26th July 2009, noted specifically as

Use curses to parse keys, which is more portable and removes the Escape delay

under the heading "Curses port."

For an example of how it was incorporated into a variant that had forked from an earlier version of Angband: PosChengband incorporated it on February 25, 2015 as "Curses Port: Fix unacceptable delay on ESC keypress" https://github.com/NickMcConnell/pos...6af5a176ec0661, with changes to main-gcu.c, util.c, and externs.h; there were also changes to Makefile, main.c, and z-config.h in that commit, but those may have been incidental.