Dosbox-staging: Different brightness levels in graphics text versus 0.74-3

[datapackrat]

Trying dosbox-staging-windows-v0.75.0-pre-613-gf2029, I noticed that text in graphics mode are a different level of brightness than the same display in 0.74-3.

side-by-side comparison

Same program, running in VESA mode 101H, text via DOS calls. Graphics are direct memory writes.

Not sure which is "correct", but wanted to bring it to your attention.

[dreamer]

I'm not sure if it's really brightness, but rather a different colour altogether. dosbox-staging seems to be using default VGA text colour (light grey) while 0.74-3 uses white in this case for some reason. Maybe it's a bug in dosbox-staging - or maybe in SVN or 0.74-3 - I can't tell now.

Can you share the program you used here? I googled for DEBUG8 executable, but couldn't find anything. With program in hand I could bisect and find out which commit exactly introduced this difference. Perhaps someone could test on a real hardware as well.

[datapackrat]

Source (and compiler) and executable: DEBUG8

1dcell.zip

Edit: Attach CHIP-8 program for debugger. To run, from the command line:
DEBUG8 1DCELL.CH8 -F
and in the debugger:
G
Use Esc to stop the debugger.

[dreamer]

I tested DEBUG8 in dosemu - it shows the same behaviour as dosbox-staging. Git bisect points to this commit as reason of different behaviour: d2209e10b5d9b4a62feb75e9dc18d81e0d7a9630 (yup it's from 2015)

It seems like this was an intentional fix, but only made in SVN trunk, never ported to any of 0.74* releases.

Closing this report; if new facts will come up (maybe someone wants to run it on real hardware?), feel free to reopen it. Thank you! :)

[kcgen]

Awesome work bisecting that out!