davidrg
commented
1 month ago Having thought about it for a few months, I think a simpler solution will be to do a bit of refactoring. At some point I'll have a go at:
- Create a new Vscrn struct that holds primary and alternate video buffers (videobuffer), a pointer to the current buffer (also videobuffer), and the vscrn type (VCMD, VTERM, etc).
- videobuffer vscrn[VNUM] in ckoco2.c would become Vscrn vscrn[VNUM]
- Everything that is currently expecting vscrn[x] to be a videobuffer is refactored to refer to vscrn[x]->currentBuffer
- Alternate screen buffer switching is achieved by changing the value of currentBuffer to point to one of the two videobuffers
- perhaps move tt_scrsize into videobuffer, or add a flag to videobuffer to indicate if its scrollable
Ideally we'd eventually move all the other values that currently live in arrays dimensioned by VNUM to values in the Vscrn struct.
Main advantages of this approach:
- Fewer changes to fewer modules (I hope)
- Fewer arrays enlarged unnecessarily compared to adding the alternate buffer as another Vscrn
- Reduction in the number of global variables
- Provides a pathway to further reduction of terminal emulation-related globals
- Provides a pathway to terminal multiplexing
- Generally tidier implementation (I hope)
If we eventually moved all of the terminal emulation-related globals into Vscrn, then we could support things like ANSI emulation in the command screen as well as multiplexed terminal sessions (TD/SMP over serial, multiple pty channels over one SSH connection, etc).
This is an option for implementing xterms Alternate Screen Buffer (#231). The solution here is to effectively add a second vscrn, VTERM_B, and rename the existing VTERM to VTERM_A.
Both Vscrns share the same event queue, and settings that are applied to one should for the most part also be applied to the other, but they have their own videobuffers allowing us to switch cleanly between them just as we switch between VTERM and VCMD. The alternate buffer, VTERM_B, has no scrollback.
These changes also introduce a new command for controlling the alternate screen buffer - it can be enabled or disabled, and there are hidden commands to force the terminal on to or of off the alternate buffer.
I'm not entirely happy with the scale of the changes here however. Adding a new Vscrn involves adding an additional member to a lot of arrays and its not always obvious which ones should be updated. Having the two VTERMs share some settings and not others, and also share their event queues (even though VTERM_B does technically have its own event queue allocated) is messy and seems a likely source for bugs.
But I don't see a simpler option. We could perhaps instead try just swapping out the contents of VTERMs videobuffer but thats going to involve similar effort keeping the two videobuffers settings in sync and adds some difficulties around disabling scrollback on the alternate buffer, but it does remove the mess around input queues. It is however probably worth investigating this approach in a separate branch just to validate whether this path is indeed the best option.
Before merging: