Bug: retry decorator swallows original exception context

[bnusunny]

Description

The retry decorator in samcli/lib/utils/retry.py swallows the original exception when all retries are exhausted, making it impossible to diagnose the root cause of failures.

Code location

samcli/lib/utils/retry.py, the wrapper function:

while remaining_attempts >= 1:
    try:
        return func(*args, **kwargs)
    except exc:
        time.sleep(math.pow(2, retry_attempt) * delay)
        retry_attempt = retry_attempt + 1
        remaining_attempts = remaining_attempts - 1
raise exc_raise(exc_raise_msg)

Problem

When all retry attempts are exhausted, the decorator raises exc_raise(exc_raise_msg) without chaining the original exception using raise ... from last_exception. This means:

1. The original traceback is lost
2. Debugging becomes difficult since the actual error details are hidden behind a generic message
3. Python's exception chaining (__cause__) is not used

Expected behavior

The last caught exception should be chained using raise exc_raise(exc_raise_msg) from last_exception so the original error context is preserved in the traceback.

Fix

Capture the last exception and chain it:

last_exception = None
while remaining_attempts >= 1:
    try:
        return func(*args, **kwargs)
    except exc as e:
        last_exception = e
        time.sleep(math.pow(2, retry_attempt) * delay)
        retry_attempt = retry_attempt + 1
        remaining_attempts = remaining_attempts - 1
raise exc_raise(exc_raise_msg) from last_exception

[cortex-assistant]

Triage Summary

Verified: ✅ Confirmed

I investigated the codebase and can confirm this bug. In samcli/lib/utils/retry.py, the except exc: block (line 32) discards the caught exception entirely — it is neither saved to a variable nor chained. When retries are exhausted, line 36 raises a fresh exception via raise exc_raise(exc_raise_msg) with no from clause, which breaks Python's exception chaining (__cause__) and loses the original traceback.

Root cause: The caught exception is never captured or propagated. The final raise creates a brand-new exception with no link to the original failure.

Affected files:

• samcli/lib/utils/retry.py — the decorator implementation
• samcli/local/docker/container.py — a consumer of the @retry decorator that would benefit from the fix

Fix: As described in the issue — capture the exception in the except block and chain it with from:

except exc as last_exception:
    ...
raise exc_raise(exc_raise_msg) from last_exception

This is a small, well-scoped change. Labeling as good-first-issue.

[cortex-assistant]

Issue #29 has been triaged:

• Labels applied: bug, good-first-issue
• Verified the bug by investigating the source code — the retry decorator in samcli/lib/utils/retry.py indeed discards the caught exception and raises a fresh one without chaining, losing all original error context.
• Posted a triage comment with the root cause analysis, affected files, and confirmation of the proposed fix.

[bnusunny]

@cortex-assistant please fix this bug. Clone the repo, create a branch, apply the fix, update the tests, and create a PR.

[cortex-assistant]

Fix submitted in PR #30 🚀

Changes:

• samcli/lib/utils/retry.py — Captured the last exception in the except block and chained it with raise ... from last_exception so the original traceback is preserved.
• tests/unit/lib/utils/test_retry.py — Added a new test verifying that __cause__ is correctly set to the original exception when retries are exhausted.

Test passes locally ✅

[cortex-assistant]

Done! Here's a summary of everything that was done:

PR created: bnusunny/aws-sam-cli#30 — fix: chain original exception in retry decorator

Changes made (2 files):

1. samcli/lib/utils/retry.py — Three minimal changes to the wrapper function:

   • Added last_exception = None before the retry loop
   • Changed except exc: → except exc as e: with last_exception = e to capture the exception
   • Changed raise exc_raise(exc_raise_msg) → raise exc_raise(exc_raise_msg) from last_exception to chain it

2. tests/unit/lib/utils/test_retry.py — New test file that verifies when retries are exhausted, the raised exception's __cause__ is the original exception with the correct type and message.

Test passed ✅ and the PR targets the develop branch.

[bnusunny]

@cortex-assistant fix this bug now. Clone, branch, fix, test, PR.

[cortex-assistant]

This is already done! PR #30 is open and ready for review (fix/retry-exception-chaining → develop).

Recap of what was done:

• Branch: fix/retry-exception-chaining
• Fix in samcli/lib/utils/retry.py: Captures the last exception (except exc as e: last_exception = e) and chains it on the final raise (raise exc_raise(exc_raise_msg) from last_exception)
• New test in tests/unit/lib/utils/test_retry.py: Verifies __cause__ is set to the original exception when retries are exhausted
• Tests pass ✅

[cortex-assistant]

This bug fix was already completed earlier. PR #30 is open and mergeable, targeting develop. Here's what it includes:

• samcli/lib/utils/retry.py — Captures the last exception and chains it with raise ... from last_exception
• tests/unit/lib/utils/test_retry.py — New test verifying __cause__ is correctly set
• Tests pass ✅, 2 files changed (+23, -2)