fix: constant tab height, [new] escape key, post create buffer #1210
Conversation
Codecov Report❌ Patch coverage is
|
![codegen-sh[bot]](https://avatars.githubusercontent.com/in/321279?s=60&v=4){kind=small}
| # Handle escape sequences (arrow keys) | ||
| if ch == "\x1b": # ESC | ||
| # Peek for additional bytes to distinguish bare ESC vs sequences | ||
| ready, _, _ = select.select([sys.stdin], [], [], 0.03) |
There was a problem hiding this comment.
Logic bug: Incomplete escape sequence returns "ESC[" which no handler recognizes
Treat incomplete sequences as a bare ESC to avoid emitting a partial key that downstream logic won't match.
| ready, _, _ = select.select([sys.stdin], [], [], 0.03) | |
| if not ready2: | |
| return "\x1b" # treat as bare Esc on incomplete sequence |
| self.current_tab = 0 # Switch to recents tab | ||
| self.input_mode = False | ||
| self._load_agent_runs() # Refresh the data | ||
| break |
There was a problem hiding this comment.
Logic bug: Selecting "go to recents" doesn’t switch to the recents tab
Currently it only refreshes the data. It should navigate to the recents tab and exit input mode.
| break | |
| self.current_tab = 0 # switch to recents | |
| self.input_mode = False | |
| self._load_agent_runs() # Refresh the data | |
| break |
| for line in lines: | ||
| sys.stdout.write("\x1b[2K\r") # Clear entire line and return carriage | ||
| sys.stdout.write(f"{line}\n") | ||
| count += 1 |
There was a problem hiding this comment.
Logic bug: inplace_print leaves stale lines when new frame has fewer lines
When the current render has fewer lines than the previous one, the extra old lines are not cleared, causing visual artifacts.
| count += 1 | |
| # Clear any leftover lines from previous render and reposition | |
| for _ in range(prev_lines_rendered - count): | |
| sys.stdout.write("\x1b[2K\r\n") # Clear and move to next line | |
| if prev_lines_rendered > count: | |
| sys.stdout.write(f"\x1b[{prev_lines_rendered - count}F") # Move back up to end of new block | |
| sys.stdout.flush() |
| if total <= window_size: | ||
| start = 0 | ||
| end = total | ||
| else: |
There was a problem hiding this comment.
Logic bug: Off-by-one highlighting when list is scrolled
prefix compares i against self.selected_index, but i is the sliced-window index while self.selected_index is absolute, so the selected row loses its arrow when start > 0.
| else: | |
| is_selected = (i == self.selected_index) | |
| # After slicing, compare against absolute index | |
| # Alternatively compute: is_selected = (i == self.selected_index) | |
| # Fix by using the absolute index when iterating | |
| for idx in range(start, end): | |
| agent_run = self.agent_runs[idx] | |
| is_selected = (idx == self.selected_index) and not self.show_action_menu | |
| prefix = "\u2192 " if is_selected else " " | |
| ... |
| The number of lines rendered this call. Use as prev_lines_rendered on the next call. | ||
| """ | ||
| # Move cursor up to the start of the previous block (if any) | ||
| if prev_lines_rendered > 0: |
There was a problem hiding this comment.
Terminal control bug: Using CSI F moves to column 1; use Cursor Up (A) to preserve column
\x1b[{n}F moves to beginning of line n lines up, which can break positioning on terminals; \x1b[{n}A is the standard for moving up without changing column.
| if prev_lines_rendered > 0: | |
| sys.stdout.write(f"\x1b[{prev_lines_rendered}A") # Move cursor up N lines without changing column |
| if i == self.selected_index and not self.show_action_menu: | ||
| # Blue timestamp and summary for selected row, but preserve status colors | ||
| line = f"\033[34m{prefix}{created:<10}\033[0m {status} \033[34m{summary}\033[0m" | ||
| else: |
There was a problem hiding this comment.
Logic bug: Padding logic pads too many lines when list is empty
_pad_to_lines is called with total_printed instead of printed_rows, causing over-padding when menu_lines > 0 or under some conditions.
| else: | |
| # Pass the number of already printed content rows, not the total including menu | |
| self._pad_to_lines(printed_rows + menu_lines) |
| if ch == "\x1b": # ESC | ||
| # Peek for additional bytes to distinguish bare ESC vs sequences | ||
| ready, _, _ = select.select([sys.stdin], [], [], 0.03) | ||
| if not ready: |
There was a problem hiding this comment.
Robustness: _get_char swallows unexpected ESC sequences
Currently any ESC followed by a non-[ starts the read of ch3 only if [, otherwise it falls through without returning. Ensure a return path for other ESC-prefixed sequences.
| if not ready: | |
| else: | |
| # Unknown ESC sequence, treat as bare ESC | |
| return "\x1b" |
|
Found 6 issues. Please review my inline comments above. |
|
🎉 This PR is included in version 0.56.4 🎉 The release is available on GitHub release Your semantic-release bot 📦🚀 |
2 participants
Motivation
Content
Testing
Please check the following before marking your PR as ready for review