[MBornes5] Graphics rendering glitch

[kapfab]

Description

Opolua does not behave like a real device for a specific graphics rendering operation.

Metadata

Key	Value
Title	MBornes5
Source URL	https://archive.org/download/3-libjune-05/3LIBJUNE05.iso/library/epocgames/millesbournes/MBornes5.SIS

Details

When the last card of the player’s hand (the one at the bottom) is discarded, it should disappear (other cards are just blanked).
That’s what happens on a real device.
With Opolua, this last card does not disappear when discarded: it's sent behind the other cards and stays visible (see screenshot).

This behaviour makes me think of gRANK or gORDER making use of an incorrect drawable position (z-index), maybe just shifted by 1, but this might be something else.

[tomsci]

Thanks for the report. Looking at the code there is a potential off-by-one error due to the toolbar clock widget, which internally is treated like another window but shouldn't be considered by the window positioning calculations. That might be the cause.

[tomsci]

Interestingly, this turns out to not be due to z-ordering but because MBorne5 specifies a mode of $101 in a call to gCREATEBIT, a value which is technically only valid when creating a window with gCREATE. Our implementation was failing to create the bitmap due to the (perceived) invalid mode but not erroring the program, meaning when it tried to clear that bit of the screen by copying from the invalid bitmap, nothing happened. Ignoring the invalid mode makes the program render correctly.

So there are several fixes to be made here:

• Ignore any flags in the mode in gCREATEBIT to match how the psion 5 (apparently) does it
• Failing to construct a bitmap in the native code should cause an error rather than silently continuing with out-of-sync info between Lua and Swift logic.
• Using an invalid bitmap id in a gCOPY call should cause a KErrDrawNotOpen error.

[kapfab]

Good news, this will probably benefit a few other programs!

[jbmorley]

@kapfab Seems like these kinds of oddities would be great to record in some OPL-level documentation. 🤔