retry failed jobs

news/197.feature

@@ -0,0 +1 @@
+Retry failed jobs when syncing packages.

pdm/core.py

@@ -109,7 +109,9 @@ def main(self, args=None, prog_name=None, obj=None, **extra):
                 etype, err, traceback = sys.exc_info()
                 if stream.verbosity > stream.NORMAL:
                     raise err.with_traceback(traceback)
-                stream.echo("[{}]: {}".format(etype.__name__, err), err=True)
+                stream.echo(
+                    f"{stream.red('[' + etype.__name__ + ']')}: {err}", err=True
+                )
                 sys.exit(1)
 
     def register_command(

pdm/installers/synchronizers.py

@@ -1,8 +1,8 @@
 import contextlib
+import functools
 import multiprocessing
 import traceback
 from concurrent.futures.thread import ThreadPoolExecutor
-from functools import partial
 from typing import Dict, List, Tuple
 
 from pip._vendor.pkg_resources import Distribution, safe_name
@@ -61,19 +61,20 @@ def __exit__(self, *args, **kwargs):
 class Synchronizer:
     """Synchronize the working set with given installation candidates"""
 
-    RETRY_TIMES = 1
     SEQUENTIAL_PACKAGES = ("pip", "setuptools", "wheel")
 
     def __init__(
         self,
         candidates: Dict[str, Candidate],
         environment: Environment,
+        retry_times: int = 1,
     ) -> None:
         self.candidates = candidates
         self.environment = environment
         self.parallel = environment.project.config["parallel_install"]
         self.all_candidates = environment.project.get_locked_candidates("__all__")
         self.working_set = environment.get_working_set()
+        self.retry_times = retry_times
 
     @contextlib.contextmanager
     def create_executor(self):
@@ -271,33 +272,37 @@ def synchronize(self, clean: bool = True, dry_run: bool = False) -> None:
                     parallel_jobs.append((kind, key))
 
         errors: List[str] = []
-        with stream.indent("  "):
-            for job in sequential_jobs:
-                kind, key = job
-                try:
-                    handlers[kind](key)
-                except Exception as err:
-                    errors.append(f"{kind} {stream.green(key)} failed:\n")
-                    errors.extend(
-                        traceback.format_exception(type(err), err, err.__traceback__)
-                    )
+        failed_jobs: List[Tuple[str, str]] = []
 
         def update_progress(future, kind, key):
             if future.exception():
-                errors.append(f"{kind} {stream.green(key)} failed:\n")
+                failed_jobs.append((kind, key))
                 error = future.exception()
                 errors.extend(
-                    traceback.format_exception(type(error), error, error.__traceback__)
+                    [f"{kind} {stream.green(key)} failed:\n"]
+                    + traceback.format_exception(
+                        type(error), error, error.__traceback__
+                    )
                 )
 
         with stream.logging("install"):
-            with stream.indent("  "), self.create_executor() as executor:
-                for job in parallel_jobs:
+            with stream.indent("  "):
+                for job in sequential_jobs:
                     kind, key = job
-                    future = executor.submit(handlers[kind], key)
-                    future.add_done_callback(
-                        partial(update_progress, kind=kind, key=key)
-                    )
+                    handlers[kind](key)
+                for i in range(self.retry_times + 1):
+                    with self.create_executor() as executor:
+                        for job in parallel_jobs:
+                            kind, key = job
+                            future = executor.submit(handlers[kind], key)
+                            future.add_done_callback(
+                                functools.partial(update_progress, kind=kind, key=key)
+                            )
+                    if not failed_jobs or i == self.retry_times:
+                        break
+                    parallel_jobs, failed_jobs = failed_jobs, []
+                    errors.clear()
+                    stream.echo("Retry failed jobs")
 
             if errors:
                 stream.echo(stream.red("\nERRORS:"))