[LLDB][Editline] empty current line before el_wgets

[imkiva]

This PR fixes #157637 by printing ANSI sequence "\r\x1b[K" to empty the line before calling to el_wgets. Later when el_wgets prints the real prompt, all wrongly printed characters are removed from the terminal.

[llvmbot]

@llvm/pr-subscribers-lldb

Author: Kiva (imkiva)

Changes

This PR fixes #157637 by printing ANSI sequence "\r\x1b[K" to empty the line before calling to el_wgets. Later when el_wgets prints the real prompt, all wrongly printed characters are removed from the terminal.

---

Full diff: https://github.com/llvm/llvm-project/pull/165830.diff

2 Files Affected:

• (modified) lldb/source/Host/common/Editline.cpp (+6)
• (modified) lldb/test/API/terminal/TestEditline.py (+1-1)

diff --git a/lldb/source/Host/common/Editline.cpp b/lldb/source/Host/common/Editline.cpp
index 1b1922e710764..03b442cce973e 100644
--- a/lldb/source/Host/common/Editline.cpp
+++ b/lldb/source/Host/common/Editline.cpp
@@ -1626,6 +1626,12 @@ bool Editline::GetLine(std::string &line, bool &interrupted) {
   m_editor_status = EditorStatus::Editing;
   m_revert_cursor_index = -1;
 
+  if (m_locked_output && m_output_stream_sp) {
+    FILE *f = m_locked_output->GetFile().GetStream();
+    fprintf(f, "\r" ANSI_CLEAR_RIGHT);
+    fflush(f);
+  }
+
   int count;
   auto input = el_wgets(m_editline, &count);
 
diff --git a/lldb/test/API/terminal/TestEditline.py b/lldb/test/API/terminal/TestEditline.py
index 38f4f34ed740b..4696b1e1b112e 100644
--- a/lldb/test/API/terminal/TestEditline.py
+++ b/lldb/test/API/terminal/TestEditline.py
@@ -94,7 +94,7 @@ def test_prompt_no_color(self):
         # after the prompt.
         self.child.send("foo")
         # Check that there are no escape codes.
-        self.child.expect(re.escape("\n(lldb) foo"))
+        self.child.expect(re.escape("\n\r\x1b[K(lldb) foo"))
 
     @skipIfAsan
     @skipIfEditlineSupportMissing

[DavidSpickett]

Thanks for looking at this! I don't know enough here to review, but I think @JDevlieghere does.

[JDevlieghere]

The flush shouldn't be necessary, the locked output streams will flush when they're unlocked.

[imkiva]

Fixed, thanks~

[JDevlieghere]

We shouldn't have to check m_output_stream_sp, and the code on 1638 assumes that we have a locked output stream, so let's replace the check with an assert here (to cover both) and then use it unconditionally like we do below.

[imkiva]

Fixed, thanks~

[JDevlieghere]

LGTM