[insjang]

The parser now handles what it does not know as DEC's terminals and xterm do (https://vt100.net/emu/dec_ansi_parser):

• an escape with intermediate bytes is dropped up to its final byte; ESC # 6 used to show "6", ESC SP F showed "F"
• DCS, SOS, PM and APC strings (ESC P, ESC X, ESC ^, ESC _) are dropped up to their ST instead of being shown, e.g. an XTGETTCAP query or a tmux passthrough
• C0 controls without a meaning here (SO and SI aside) and DEL are not written as characters; VT and FF move down as LF does
• inside a sequence, CAN and SUB cancel it, ESC starts a new one, and the other controls run in place where they used to become part of a parameter
• a control sequence with an intermediate byte, or with a private marker other than '?' and the '>' of DA2, is dropped instead of being run as another command: CSI 1 SP @ is not ICH, CSI = 1049 h does not switch to the alternate screen

Commands and modes added:

• IND, NEL, HTS (ESC D, ESC E, ESC H), TBC (CSI g), CHT and CBT (CSI I, CSI Z); tab stops start every 8 columns
• HVP (CSI f), REP (CSI b), SCOSC and SCORC (CSI s, CSI u)
• DA1 answers a VT100 with advanced video (CSI ? 1 ; 2 c), as xterm in VT100 mode; DA2 answers CSI > 0 ; 0 ; 0 c
• DECALN (ESC # 8) fills the screen with E's in the default rendition, resets the margins and homes the cursor
• DEC special graphics (ESC ( 0, ESC ) 0, SO, SI) draw the line drawing characters that TERM=xterm promises with smacs/acsc; the national sets are taken as ASCII
• DECAWM off (CSI ? 7 l): a character at the right margin replaces the last one and the cursor stays
• DECOM (CSI ? 6 h): cursor positions and CPR count from the top margin and stay inside the scroll region

Fixed on the way:

• a count or position of 0 means 1 (ECMA-48), for the cursor moves, ICH, DCH, ECH, IL, DL, SU and SD; the backend still takes a count of 0 as nothing
• DECSTBM homes the cursor, as on a VT100
• CUU, CUD, CNL and CPL stop at the margins from inside the scroll region
• IL, DL, SU and SD with a count past the scroll region empty it and leave the lines outside it alone; they used to clear lines below the region, and above it when IL and DL followed each other
• ESC 7 saves the rendition, the origin mode and the character sets with the cursor, as DECSC; a full reset forgets what was saved

Not covered: 8-bit controls (S7C1T, S8C1T), VT52 mode, 132 columns (DECCOLM), double width and height lines, protected areas (DECSCA), DECSTR and DECRQM.

Tests: 23 new cases in VT100EmulatorTest, each failing without the change it covers.

vttest 2.7 (menus 1-4 and 6-11, keyboard and mouse left out), with the wrap deferred as the local, SSH and Telnet connectors set it, compared screen by screen with tmux running the same keys: 370 of the 390 screens both runs reach match, 307 on master. Of the screens that matched on master, the two that differ now show the line drawing characters, which tmux's capture prints as the letters it was sent.

How to test: run vttest (e.g. apt install vttest) in a Terminal, and go through menu 1 (cursor movements), 2 (screen features), 3 (character sets) and 8 (insert/delete). On master the E frame of menu 1 is missing, digits of ESC # show up, and the line drawing set shows letters.

This branch is on master. It touches the emulator as #2901, #2902, #2903 and #2904 do, so it will need a rebase as those are merged; I will do that.

[akurtakov]

I would appreciate if you slow down a bit opening PRs as there are way too many notifications coming in from rebasing all terminal PRs when one lands in. In an ideal world that should lure more reviewers but I truly doubt this will happen.

[github-actions]

Test Results

    54 files  ±  0      54 suites  ±0   34m 1s ⏱️ - 4m 32s
 5 039 tests + 35   5 017 ✅ + 35   22 💤 ±0  0 ❌ ±0 
13 008 runs  +105  12 854 ✅ +105  154 💤 ±0  0 ❌ ±0 

Results for commit f7eb418. ± Comparison against base commit bc125e3.

♻️ This comment has been updated with latest results.

[insjang]

Sorry about the noise, and thanks for saying so. I'll stop opening new PRs until the open ones are through, and I won't rebase the others whenever one lands: I'll rebase a PR only when you are about to look at it or ask for it, and push only to answer a review, with the fixes for one review in one push. #3059 can wait until the rest are done. I'll run Copilot on separate branches in my fork first, so those runs don't notify anyone here.

[Copilot]

🟡 Changes recommended

Unresolved terminal-emulation defects cause incorrect wrapping, stale character-set state, incomplete cursor-state restoration, unintended screen clearing, and truncated repeated output.

6 open issues

• Save character sets and origin mode in the 1049 snapshot · New
• Clear stale line-drawing state after charset designation · New
• Require exactly ESC # 8 for DECALN · New
• Clear saved wrap state during reset · New
• Honor REP counts beyond one screen of output · New
• Clamp tab destinations to the current right margin · New

What changed in this PR

This PR improves Eclipse Terminal’s VT100 and ECMA-48 compatibility by tightening escape-sequence parsing and expanding support for terminal commands exercised by vttest.

Changes:

• Handles unsupported sequences and embedded controls without displaying stray characters.
• Adds tab stops, line drawing, device reports, cursor modes, and related commands; corrects scrolling and wrapping behavior.
• Adds regression tests and captures terminal replies in the test mock.

File	Description
terminal/​tests/​org.eclipse.terminal.test/​src/​org/​eclipse/​terminal/​internal/​emulator/​VT100EmulatorTest.java	Adds parser, command, mode, and screen-behavior regression tests.
terminal/​tests/​org.eclipse.terminal.test/​src/​org/​eclipse/​terminal/​internal/​emulator/​MockTerminalControlForText.java	Captures terminal replies for assertions.
terminal/​bundles/​org.eclipse.terminal.control/​src/​org/​eclipse/​terminal/​internal/​emulator/​VT100EmulatorBackend.java	Implements auto-wrap control, screen filling, and bounded region scrolling.
terminal/​bundles/​org.eclipse.terminal.control/​src/​org/​eclipse/​terminal/​internal/​emulator/​VT100Emulator.java	Expands parsing, command handling, character sets, and saved state.
terminal/​bundles/​org.eclipse.terminal.control/​src/​org/​eclipse/​terminal/​internal/​emulator/​VT100BackendTraceDecorator.java	Delegates and traces the new backend operations.
terminal/​bundles/​org.eclipse.terminal.control/​src/​org/​eclipse/​terminal/​internal/​emulator/​IVT100EmulatorBackend.java	Defines backend operations needed by the new terminal behavior.

_{Review effort: Balanced}