SPMI: More information on CI failures (#77361)

This enhances parts of the SPMI process with more information on failures.

Also fixes a bug I introduced when I added tpdiffs: when asmdiffs fails, it was not marking the entire pipeline as failing.

src/coreclr/scripts/superpmi.py

@@ -1691,10 +1691,7 @@ async def create_replay_artifacts(print_prefix, context_index, self, mch_file, e
                         """
                         # Setup flags to call SuperPMI for both the diff jit and the base jit
 
-                        flags = [
-                            "-c", str(context_index),
-                            "-v", "q"  # only log from the jit.
-                        ]
+                        flags = ["-c", str(context_index)]
                         flags += altjit_replay_flags
 
                         # Change the working directory to the core root we will call SuperPMI from.
@@ -1709,11 +1706,21 @@ async def create_one_artifact(jit_path: str, location: str, flags) -> str:
                                 modified_env['DOTNET_JitStdOutFile'] = item_path
                                 logging.debug("%sGenerating %s", print_prefix, item_path)
                                 logging.debug("%sInvoking: %s", print_prefix, " ".join(command))
-                                proc = await asyncio.create_subprocess_shell(" ".join(command), stderr=asyncio.subprocess.PIPE, env=modified_env)
-                                await proc.communicate()
-                                with open(item_path, 'r') as file_handle:
-                                    generated_txt = file_handle.read()
-                                return generated_txt
+                                proc = await asyncio.create_subprocess_shell(" ".join(command), stdout=asyncio.subprocess.PIPE, stderr=asyncio.subprocess.PIPE, env=modified_env)
+                                (stdout, stderr) = await proc.communicate()
+
+                                def create_exception():
+                                    return Exception("Failure while creating JitStdOutFile.\nExit code: {}\nstdout:\n{}\n\nstderr:\n{}".format(proc.returncode, stdout.decode(), stderr.decode()))
+
+                                if proc.returncode != 0:
+                                    # No miss/replay failure is expected in contexts that were reported as having diffs since then they succeeded during the diffs run.
+                                    raise create_exception()
+
+                                try:
+                                    with open(item_path, 'r') as file_handle:
+                                        return file_handle.read()
+                                except BaseException as err:
+                                    raise create_exception() from err
 
                             # Generate diff and base JIT dumps
                             base_txt = await create_one_artifact(self.base_jit_path, base_location, flags + base_option_flags_for_diff_artifact)

src/coreclr/scripts/superpmi_diffs.py

@@ -218,6 +218,11 @@ def main(main_args):
         "-log_level", "debug",
         "-log_file", log_file])
 
+    failed = False
+    if return_code != 0:
+        print("Failed during asmdiffs")
+        failed = True
+
     print("Running superpmi.py tpdiff")
 
     overall_md_tpdiff_summary_file = os.path.join(spmi_location, "tpdiff_summary.md")
@@ -240,6 +245,10 @@ def main(main_args):
         "-log_level", "debug",
         "-log_file", log_file])
 
+    if return_code != 0:
+        print("Failed during tpdiff")
+        failed = True
+
     # If there are asm diffs, and jit-analyze ran, we'll get a diff_summary.md file in the spmi_location directory.
     # We make sure the file doesn't exist before we run diffs, so we don't need to worry about superpmi.py creating
     # a unique, numbered file. If there are no diffs, we still want to create this file and indicate there were no diffs.
@@ -269,7 +278,7 @@ def main(main_args):
     # Finally prepare files to upload from helix.
     copy_dasm_files(spmi_location, log_directory, "{}_{}".format(platform_name, arch_name))
 
-    if return_code != 0:
+    if failed:
         print("Failure in {}".format(log_file))
         return 1