fix(): S3 compat for on-prem (s3proxy/MinIO) and graceful empty-log handling

[MatthewSwords]

Summary

• S3 checksum compatibility: When s3_endpoint_url is set (Nexus/on-prem deployments), disable default boto3 checksum headers (x-amz-checksum-*) via BotocoreConfig(request_checksum_calculation="when_required"). s3proxy does not implement these headers, causing PutObject to fail with NotImplemented. This is a no-op for production (AWS S3) since the setting only activates when a custom endpoint is configured.
• Graceful empty-log handling: case.logs.tail() on a job that never produced solver logs (e.g. meshing failed before solver ran) threw IndexError from an unchecked [0] on an empty list. Now raises FileNotFoundError with a clear message.

Context

Discovered during Nexus dev setup testing with the flow360 Python client. The S3 issue affects all file uploads when using s3proxy (Nexus dev) or certain MinIO versions (Nexus release packages). The log fix is a minor edge case for errored jobs.

Test plan

• Verified S3 upload works against s3proxy with this fix (geometry upload succeeds)
• Verified case._download_file() still works for file downloads
• Verified case.logs.tail() now gives FileNotFoundError instead of IndexError when no logs exist
• Verify production (AWS S3) uploads still work (no behavior change when s3_endpoint_url is None)

Made with Cursor

---

Note

Medium Risk
Changes S3 client configuration (checksums and addressing style) when using a custom endpoint, which could affect upload/download behavior in on-prem deployments. Also changes log retrieval to raise a clearer runtime error when no log files exist, altering error behavior for some failed jobs.

Overview
Improves S3 compatibility for on-prem/S3-compatible backends by customizing BotocoreConfig when Env.current.s3_endpoint_url is set: disables default checksum headers, forces path-style addressing, and falls back gracefully on older botocore versions.

Updates RemoteResourceLogs._get_tmp_file_name() to detect when no .log files are available and raise Flow360RuntimeError with a clear message instead of failing via list indexing; adds tests covering empty and non-matching file lists.

^{Written by Cursor Bugbot for commit c5852ee. This will update automatically on new commits. Configure here.}

[cursor]

Cursor Bugbot has reviewed your changes and found 1 potential issue.

^{Bugbot Autofix is OFF. To automatically fix reported issues with cloud agents, enable autofix in the Cursor dashboard.}

[Copilot]

Pull request overview

This PR improves the Flow360 Python client’s robustness in two areas that commonly affect on-prem/Nexus dev setups: S3-compatible storage uploads (e.g., s3proxy/MinIO) and handling resources that never produced solver logs.

Changes:

• Adjust S3 client construction to disable default checksum header behavior when using a custom s3_endpoint_url (S3-compatible backends).
• Make RemoteResourceLogs raise a clear FileNotFoundError when no log files exist instead of failing with an IndexError.

Reviewed changes

Copilot reviewed 2 out of 2 changed files in this pull request and generated 3 comments.

File	Description
flow360/component/resource_base.py	Adds explicit empty-log detection and raises FileNotFoundError with a clearer message.
flow360/cloud/s3_utils.py	Extends botocore client config for custom S3 endpoints to avoid incompatible checksum headers/validation.

Comments suppressed due to low confidence (1)

flow360/cloud/s3_utils.py:169

• There’s now behavior branching on Env.current.s3_endpoint_url to change checksum settings, but there are existing tests for flow360/cloud/s3_utils.py (e.g. tests/test_s3_utils.py) and none validate that get_client() passes the intended BotocoreConfig when a custom endpoint is set. Adding a unit test that monkeypatches boto3.client to capture the config argument would help prevent regressions in the on-prem compatibility fix.

        config_kwargs = {"max_pool_connections": MAX_POOL}
        if Env.current.s3_endpoint_url is not None:
            # S3-compatible stores (s3proxy, MinIO) may not implement the
            # checksum headers that newer boto3 versions send by default.
            config_kwargs["request_checksum_calculation"] = "when_required"
            config_kwargs["response_checksum_validation"] = "when_required"

        kwargs = {
            "region_name": self.user_credential.region,
            "aws_access_key_id": self.user_credential.access_key_id,
            "aws_secret_access_key": self.user_credential.secret_access_key,
            "aws_session_token": self.user_credential.session_token,
            "config": BotocoreConfig(**config_kwargs),
        }

        if Env.current.s3_endpoint_url is not None:
            kwargs["endpoint_url"] = Env.current.s3_endpoint_url

        return client("s3", **kwargs)

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[mikeparkflex]

My agent with context from debugging concurs. No open items remaining. Both PRs are ready to merge.