Chunk very long commands when writing file. (#247)

* Chunk very long commands when writing file.

* Split tests in smaller ones

Author: MarcCote

debug_gym/gym/workspace.py

@@ -136,15 +136,25 @@ def write_file(self, filepath: str, content: str):
         """Writes `content` to `filepath` exactly as-is, preserving any trailing newlines."""
         abs_filepath = self.resolve_path(filepath)
 
+        # We will split content in chunks of 32kB to avoid hitting command length limits.
+        chunk_size = 32 * 1024  # 32kB
+        first_chunk = content[:chunk_size]
+        rest = content[chunk_size:]
+
         # In the following command we:
         # - use a single-quoted heredoc (cat <<'nDEBUGGYM_EOF' ... nDEBUGGYM_EOF) so the heredoc body is taken literally (no shell expansion)
         # - append a sentinel character DEBUGGYM_DEL inside the heredoc so we can detect/restore trailing newlines later
         # - capture the heredoc output into shell variable CONTENT since command substitution strips trailing newlines
         # - "${CONTENT%DEBUGGYM_DEL}" removes the trailing sentinel DEBUGGYM_DEL (restoring the original trailing-newline state)
         # - echo -n writes the result without adding an extra newline
-        cmd = f"CONTENT=$(cat <<'DEBUGGYM_EOF'\n{content}DEBUGGYM_DEL\nDEBUGGYM_EOF\n); echo -n \"${{CONTENT%DEBUGGYM_DEL}}\" > {abs_filepath}"
+        cmd = f"CONTENT=$(cat <<'DEBUGGYM_EOF'\n{first_chunk}DEBUGGYM_DEL\nDEBUGGYM_EOF\n); echo -n \"${{CONTENT%DEBUGGYM_DEL}}\" > {abs_filepath}"
         self.terminal.run(cmd, raises=True)
 
+        for i in range(0, len(rest), chunk_size):
+            chunk = rest[i : i + chunk_size]
+            cmd = f"CONTENT=$(cat <<'DEBUGGYM_EOF'\n{chunk}DEBUGGYM_DEL\nDEBUGGYM_EOF\n); echo -n \"${{CONTENT%DEBUGGYM_DEL}}\" >> {abs_filepath}"
+            self.terminal.run(cmd, raises=True)
+
     def directory_tree(self, root: str | Path = None, max_depth: int = 1):
         root = self.resolve_path(root or self.working_dir, raises=True)
         # Use the terminal to run a bash command to list files

debug_gym/logger.py

@@ -518,7 +518,12 @@ def __init__(
     ):
         super().__init__(name)
         # If var env "DEBUG_GYM_DEBUG" is set, turn on debug mode
-        if os.environ.get("DEBUG_GYM_DEBUG"):
+        if os.environ.get("DEBUG_GYM_DEBUG", "").strip().lower() in (
+            "1",
+            "true",
+            "yes",
+            "on",
+        ):
             level = logging.DEBUG
 
         # Prevent the log messages from being propagated to the root logger

tests/gym/test_workspace.py

@@ -292,23 +292,43 @@ def test_read_file_raises_for_nonexistent_file(workspace):
         workspace.read_file("/test.txt")
 
 
-def test_write_file(workspace):
+def test_write_file_basic(workspace):
     file_path = workspace.working_dir / "test.txt"
     file_content = "Hello, DebugGym!\n\n\n"
     workspace.write_file("test.txt", file_content)
     assert file_path.read_text() == file_content
 
-    # Test with single line, no newline at the end.
+
+def test_write_file_single_line_no_newline(workspace):
+    file_path = workspace.working_dir / "test.txt"
     file_content_single_line = "Hello, DebugGym!"
     workspace.write_file("test.txt", file_content_single_line)
     assert file_path.read_text() == file_content_single_line
 
-    # Should still work if the content ends with the delimiter.
+
+def test_write_file_with_delimiter(workspace):
+    file_path = workspace.working_dir / "test.txt"
     file_content_single_line = "Hello, DebugGym!nDEBUGGYM_DEL"
     workspace.write_file("test.txt", file_content_single_line)
     assert file_path.read_text() == file_content_single_line
 
-    # Test with newlines
+
+def test_write_file_with_newlines(workspace):
+    file_path = workspace.working_dir / "test.txt"
     file_content_with_newlines = "Hello, DebugGym!\nThis is a test.\n"
     workspace.write_file("test.txt", file_content_with_newlines)
     assert file_path.read_text() == file_content_with_newlines
+
+
+def test_write_file_empty_content(workspace):
+    file_path = workspace.working_dir / "test.txt"
+    file_content_empty = ""
+    workspace.write_file("test.txt", file_content_empty)
+    assert file_path.read_text() == file_content_empty
+
+
+def test_write_file_exceeding_max_command_length(workspace):
+    file_path = workspace.working_dir / "test.txt"
+    file_content_exceeding_max_command_length = "A" * (2 * 1024**2)  # 2MB of 'A's
+    workspace.write_file("test.txt", file_content_exceeding_max_command_length)
+    assert file_path.read_text() == file_content_exceeding_max_command_length