gh-127068: default REPL where prompts containing newlines would be dup

[yihong0618]

The root cause of the this lis the calc_screen() method of reader.py. When screen, screeninfo, and line_end_offsets are retrieved from the cache, they are passed by direct reference instead of by a copy.

before this patch

2025-10-01.12.31.25.mov

after this patch

2025-10-01.12.57.17.mov

• Issue: New REPL repeats prompt with newline after each key #127068

[kemingy]

It seems this last_refresh_line_end_offsets has never been used in this function. The bug can be fixed by deleting this one.

The other two screen & screeninfo are synced back to self.last_refresh_cache. But for readability & maintenance, we should copy them here.

[yihong0618]

addressed

[kemingy]

The failed Windows test (test.test_pyrepl.test_windows_console.WindowsConsoleTests.test_multiline_ctrl_z) is not caused by this PR.

If we add the reader.screeninfo to the assert msg:

- (4, 2)
+ (2, 3) : [(0, [1, 1, 1, 1, 1]), (0, [1, 1, 1, 1, 1]), (0, [1, 2, 1])]

[yihong0618]

@picnixz friendly ping and also fix #142078

[TotalyEnglizLitrate]

Other escape characters being present in sys.ps1 (or indirectly via input) also cause the following issues not seen in the basic REPL

Namely (appended them at the end of >>> for demonstration)
\t - causes prompt to be shown doubly upon very first character input

\b (and/or) \r - causes displayed cursor position to be out-of-sync with true cursor position (the displayed cursor does not account for the backward shift caused by the escape sequences)

• for >>>\b - cursor is displayed 2 spaces beyond true position (1 because it perceives \b as a single char that exists after >>> and the other due to the cursor being moved 1 space back)
• for >>>\r - similar problem, actual cursor is displayed 4 spaces later for the same reason

\v - similar issue to \n - additionally causes cursor position discrepancy of a single space

there may be others I didn't catch as well

output.mp4

all of these still occur in 3.13, 3.14, main as well as in this patched version (and none occur when using the basic repl)

will/should these be addressed here or should I create a separate issue?