Display glitch with GCU client?

[PowerWyrm]

In order to play with Necklace of the Eye with my variant, I've compiled a GCU client using PDCURSES 3.4 (I'm on Win7)... but I face some weird display glitches.

Below is an example of what's happening. When the character moves from the left side of the room to the right side, the walls are incorrectly redrawn until I press ctrl-R to redraw the screen. It seems that every single character that is drawn is one pixel too large.

I'm wondering... Since the client only displays ASCII symbols using Term_text_gcu(), this problem shouldn't happen unless there's some obscure bug in the curses library. Can anyone tell me if there's the same problem with V?

[d_m]

Originally Posted by PowerWyrm

I'm wondering... Since the client only displays ASCII symbols using Term_text_gcu(), this problem shouldn't happen unless there's some obscure bug in the curses library. Can anyone tell me if there's the same problem with V?

It may be a problem with wide text (unicode) on Windows, which is UCS-2/UTF-16. That would be a bit ugly to diagnose.

It may also be a problem with using reverse video'd spaces for walls (e.g. using A_REVERSE). You could test this by changing granite walls from:

F:56:all:0x81:0x20

into:

F:56:all:0x01:0x23

Other than those two things, I'm not sure what it could be besides a bug in the curses library.

[PowerWyrm]

Not a problem with Unicode, I've tried using the non-Unicode functions to display the characters and I still have the problem.

Not a problem with walls, it's still happening when using the basic # character to display walls.

Then it's a bug in the curses library... dammit.

[d_m]

Originally Posted by PowerWyrm

Then it's a bug in the curses library... dammit.

Ouch

Well, I know that a windows console mode used to work, so I think there must be some library out there that can do it. May require some serious #ifdef stuff though...

Anyway, good luck!

[PowerWyrm]

The problem arises when using colors. Printing everything in white using wattrset(window, 0) doesn't produce the glitch. But as soon as you use wattrset(window, somecolor), the glitch happens.

I'll try to figure out if something in the current main-gcu code produces the problem, or if it comes from pdcurses. If it's a bug with pdcurses, the only other option is to download ncurses 5.9 and try the new Windows Console port that can be compiled with MinGW.

[d_m]

Originally Posted by PowerWyrm

The problem arises when using colors. Printing everything in white using wattrset(window, 0) doesn't produce the glitch. But as soon as you use wattrset(window, somecolor), the glitch happens.

I'll try to figure out if something in the current main-gcu code produces the problem, or if it comes from pdcurses. If it's a bug with pdcurses, the only other option is to download ncurses 5.9 and try the new Windows Console port that can be compiled with MinGW.

The other thing I could imagine is that maybe pdcurses is trying to support 88 or 256 colors and failing? Or that it purports to be able to reset colors? In those cases, hard-disabling those features might also help.

[PowerWyrm]

This is clearly a bug with PDCurses, as someone else reported the same problem on pdcurses's sourceforge bug tracker.

Today I've tried something really challenging... recompiling the GCU client using NCURSES... under Windows! With the latest 5.9 version, there is a Windows build that can be compiled using minGW. Unfortunately I don't have my source code configured for minGW, so I had to configure ncurses using minGW to get the headers and defines, plug the resulting ncurses source code into my BCC project, fix the few compile errors due to not having minGW headers (sys/time.h and such) and then adapt main-gcu. But it works!

The only problem here, as stated in the ncurses readme file, is that ncurses 5.9 has no widechar support when compiled under Windows. This means no fancy walls...

[Nick]

Cool. Now we can play in terminals on every platform and ditch all that funky platform specific code

[d_m]

Originally Posted by Nick

Cool. Now we can play in terminals on every platform and ditch all that funky platform specific code

You said it!

[PowerWyrm]

Originally Posted by PowerWyrm

The only problem here, as stated in the ncurses readme file, is that ncurses 5.9 has no widechar support when compiled under Windows. This means no fancy walls...

Alright... I've regenerated the minGW configuration using --enable-widec to add w_char support (mvwaddnwstr and co.). Now it's time to see if I can plug the fancy walls back...