Experimental solution for step through interpreting the full flow issue #16351 #17308
Conversation
A change in SDL2 (libsdl-org/SDL@61cd57d) caused modifier changes to be sent as a repeat keydown with different key modifiers. This causes some keybindings to become un-typeable (For example: Cmd+Shift+W, E) because the keyup events cause these new keydown repeats with the new modifiers (for the previous example these are something like Cmd+Shift/Cmd+W/Shift+W).
…out repeat events. This feels like a hack...
- fix undo/redo cases that broke text, specially when in collaboration with the syntax highlighter - make undo/redo work on a word-by-word basis - make undo/redo restore selection and cursur when doing a replacement by paste/completion Co-authored-by: Fede Lochbaum <[email protected]>
…ugSession. These tests should be parametrized in the future. They are commited like this to ease the performance comparison between DebugSession and EnhancedDebugSession
Improvements in undo/redo
…vents Fix repeat events
Remove unnecessary KM repository reset
Fix typo: change "The cyclomatic complexity is heigh" to "The cyclomatic complexity is high"
…s-editor Major update of the Keymap Descriptions to a Shortcuts Editor
…t raise an exception in the interpreter but put a new context in it with the message send taken from the Smalltalk Special Array
|
Hi, this is super interesting work! but quite complex. Also a dumb question: as far as I understand (but may be wrong) the only interest of the step through is stepping into blocks? Would it be possible to, instead of stepping, detect that the user attempts to step in to a block (by using step through near a block), then put an unknown bytecode at the start of the block and just proceed, and get back the debugger when hitting that unknown bytecode? Probably is possible that it never gets into the block and so another "security unknown bytecode" should be put after the block to catch that we just never went into. Maybe it would not work or would be more complicated that the current attempt. But I would be interested to reflect on all that can be done with the unknown bytecode trick. |
Please do not review yet
This is an experimental solution aimed at addressing issue#16351.
In it, an unknown bytecode is used to execute step through.
This is a work in progress, results and further steps will be analized with @tesonep
For performance comparison use the following code snippet