- made continuing the chat when the generated test fails or if it

  doesn't improve coverage optional, so it can be up to each Prompter;

- disabled these in gpt-v2-ablated Prompter to facilitate evaluation;

src/coverup/coverup.py

@@ -517,7 +517,11 @@ async def improve_coverage(chatter: llm.Chatter, prompter: prompt.Prompter, seg:
 
         except subprocess.CalledProcessError as e:
             state.inc_counter('F')
-            prompts = prompter.error_prompt(seg, clean_error(str(e.stdout, 'UTF-8', errors='ignore')))
+            error = clean_error(str(e.stdout, 'UTF-8', errors='ignore'))
+            if not (prompts := prompter.error_prompt(seg, error)):
+                log_write(seg, "Test failed:\n\n" + error)
+                break
+
             messages.extend(prompts)
             continue
 
@@ -538,7 +542,10 @@ async def improve_coverage(chatter: llm.Chatter, prompter: prompt.Prompter, seg:
 
         if not gained_lines and not gained_branches:
             state.inc_counter('U')
-            prompts = prompter.missing_coverage_prompt(seg, seg.missing_lines, seg.missing_branches)
+            if not (prompts := prompter.missing_coverage_prompt(seg, seg.missing_lines, seg.missing_branches)):
+                log_write(seg, "Test doesn't improve coverage")
+                break
+
             messages.extend(prompts)
             continue
 

src/coverup/logreader.py

@@ -3,7 +3,7 @@
 from pathlib import Path
 
 
-TERMINAL_EVENTS=('G', 'M', 'T', '-', '*')
+TERMINAL_EVENTS=('G', 'M', 'T', '-', '*', 'f', 'u')
 
 
 def is_same_as_P(content, begin, end):
@@ -107,12 +107,16 @@ def what(content: str):
                 return 'C'
             elif content.startswith("Executing the test yields an error"):
                 return 'F'
+            elif content.startswith("Test failed"):
+                return 'f'
             elif content.startswith("Executing the test along with"): # side effect
                 return 'S'
             elif content.startswith("```python"): # response
                 return 'R'
             elif content.startswith("This test still lacks coverage"):
                 return 'U'
+            elif content.startswith("Test doesn't improve coverage"):
+                return 'u'
             elif content.startswith("Saved as"): # success
                 return 'G'
             elif content.startswith("Missing modules"):

src/coverup/prompt.py

@@ -27,13 +27,13 @@ def initial_prompt(self, segment: CodeSegment) -> T.List[dict]:
 
 
     @abc.abstractmethod
-    def error_prompt(self, segment: CodeSegment, error: str) -> T.List[dict]:
+    def error_prompt(self, segment: CodeSegment, error: str) -> T.List[dict] | None:
         """Returns prompts(s) in response to an error."""
 
 
     @abc.abstractmethod
     def missing_coverage_prompt(self, segment: CodeSegment,
-                                missing_lines: set, missing_branches: set) -> T.List[dict]:
+                                missing_lines: set, missing_branches: set) -> T.List[dict] | None:
         """Returns prompts(s) in response to the suggested test(s) lacking coverage."""
 
 
@@ -218,21 +218,23 @@ def initial_prompt(self, segment: CodeSegment) -> T.List[dict]:
 """)
         ]
 
-    def error_prompt(self, segment: CodeSegment, error: str) -> T.List[dict]:
-        return [_message(f"""\
-Executing the test yields an error, shown below.
-Modify the test to correct it; respond only with the complete Python code in backticks.
-
-{error}""")
-        ]
+    def error_prompt(self, segment: CodeSegment, error: str) -> T.List[dict] | None:
+        return None
+#        return [_message(f"""\
+#Executing the test yields an error, shown below.
+#Modify the test to correct it; respond only with the complete Python code in backticks.
+#
+#{error}""")
+#        ]
 
     def missing_coverage_prompt(self, segment: CodeSegment,
-                                missing_lines: set, missing_branches: set) -> T.List[dict]:
-        return [_message(f"""\
-The tests still lack coverage.
-Modify to correct that; respond only with the complete Python code in backticks.
-""")
-        ]
+                                missing_lines: set, missing_branches: set) -> T.List[dict] | None:
+        return None
+#        return [_message(f"""\
+#The tests still lack coverage.
+#Modify to correct that; respond only with the complete Python code in backticks.
+#""")
+#        ]
 
 
 class ClaudePrompter(Prompter):