Enable event-driven Lambda execution for S3 Python index generation by araravik-psd · Pull Request #3737 · ROCm/TheRock

araravik-psd · 2026-03-03T23:17:37Z

This PR refactors the S3 index generation logic to support event-driven execution via AWS Lambda, in addition to existing CLI/CI workflows.

Key Changes

Refactored manage.py to expose reusable index update logic.
Regenerate indexes dynamically based on uploaded/deleted object prefix.
Removed reliance on static PREFIXES for Lambda execution.
Made S3_BUCKET_PY environment variable mandatory for safer bucket targeting.
Improved Lambda observability by returning processed prefix + object count.
Ensured recursion protection by skipping index.html events.
Support for .whl, .zip, and .tar.gz artifacts.
Updated Package selection including changes from Multi-arch package structure

Result

Clear separation between indexing logic and infrastructure.
Cleaner and more maintainable long term.
Reusable logic for both CI and server-side execution.
Enables S3 retention policy compatibility (index regenerates on object deletion).
Cleaner and more maintainable long term.

The Lambda wrapper PR is here:

https://github.com/ROCm/TheRock-Infra/pull/107/changes

Test results for newly added auto-detection logic:

https://gist.github.com/araravik-psd/4f2869e071ea1ee384cd3587f31a6660

araravik-psd · 2026-03-03T23:19:16Z

@amd-shiraz Can you point me to the location I could use in the infra repo to save a copy of the lambda function wrapper used here?

CC: @marbre @HereThereBeDragons

amd-shiraz · 2026-03-04T00:19:25Z

@amd-shiraz Can you point me to the location I could use in the infra repo to save a copy of the lambda function wrapper used here?

CC: @marbre @HereThereBeDragons

pls put the scripts inside tools: https://github.com/ROCm/TheRock-Infra/tree/main/tools (create a folder perhaps if you think multiple scripts can go in for this). for terraform related lambda code put it under projects https://github.com/ROCm/TheRock-Infra/tree/main/projects, base modules goes under modules https://github.com/ROCm/TheRock-Infra/tree/main/modules/aws .

HereThereBeDragons

as you are already touching the file it would be good to do a proper freshening up of the file and improve the overall QoL when using it :)

HereThereBeDragons · 2026-03-04T11:22:32Z

 # bucket for TheRock
 # We also manage `therock-nightly-python` (not the default to make the script safer to test)
-BUCKET_NAME = getenv("S3_BUCKET_PY", "therock-dev-python")
+BUCKET_NAME = getenv("S3_BUCKET_PY")


as this is a bit of cleanup i personally would appreciate if this could also be supported by a cli arg --bucket?

unless @ScottTodd has a different opinion?

This has its origin in upstream script. It depends on how the Lambda wrapper script will handle it an pass argument. In general I prefer to pass explicit arguments instead of using environment variables. However, can be a follow up also but command line args are preferred for manual use.

This would however mean to update our workflows as well. I'd rather do it in a follow up therefore once we switched over. Less refactoring (as calling the script in workflows will be dropped anyway).

Good points. I added --bucket, but kept S3_BUCKET_PY as a fallback so we don’t break any existing workflows or Lambda usage.

Right now it works like this:

If --bucket is passed, we use that (better for manual runs)

Otherwise we fall back to S3_BUCKET_PY

If neither is set, it fails fast

I also didn’t hardcode a default bucket on purpose — didn’t want to risk accidentally writing to prod.

Could you have a quick look at the initialize_bucket() logic and confirm this makes sense? Mainly want to make sure it’s fine from your side and doesn’t cause any surprises.

marbre · 2026-03-09T23:28:20Z

+    # TODO: bucket mirror just to hold index used with CDN
+    # BUCKET_CDN = S3.Bucket('therock-nightly-python-testing')


Suggested change

# TODO: bucket mirror just to hold index used with CDN

# BUCKET_CDN = S3.Bucket('therock-nightly-python-testing')

We don't have this.

marbre

Some additional comments from my agent:

Critical Issues

1. Missing `lambda_function.py` — the PR's primary goal is absent

The PR description says "Introduced a thin lambda_function.py wrapper to
handle S3 events" but the diff only modifies manage.py. The Lambda
integration layer never ships. Without it, this PR is a pure CLI refactor
with no Lambda functionality delivered.

2. `with_metadata` behavior regression — SHA256 checksums dropped from normal uploads

Before:

idx = S3Index.from_S3(prefix=prefix, with_metadata=True or args.compute_sha256)
# Note: `True or X` always evaluates True — metadata always fetched

After (update_pep503_index):

idx = S3Index.from_S3(prefix=prefix, with_metadata=compute_sha256)

Metadata (SHA256 checksums) is now only fetched when --compute-sha256
is passed. For normal upload runs, compute_sha256=False, so
with_metadata=False, meaning uploaded index.html files will have no
#sha256=... fragments. This silently breaks PEP 503 checksum links in the
package index.

The original True or args.compute_sha256 was itself a bug (always True),
but the fix should be with_metadata=True for normal upload paths, not
with_metadata=compute_sha256.

3. Uninitialized globals — Lambda invocations will crash with `NameError`

The original module-level assignments:

BUCKET_NAME = getenv("S3_BUCKET_PY", "therock-dev-python")
BUCKET = S3.Bucket(BUCKET_NAME)
INDEX_BUCKETS = {BUCKET}

...are removed and replaced with initialize_bucket(), which is only called
from main(). But BUCKET_NAME, BUCKET, and INDEX_BUCKETS are no longer
declared at module scope at all. Any code path (Lambda handler, import-time
side effects) that uses these before initialize_bucket() runs will get
NameError: name 'BUCKET_NAME' is not defined. The globals should at least
be declared as None at module level.

Moderate Issues

4. Loss of `prefix` validation / `all` behavior

Before: choices=PREFIXES + ["all"] enforced known prefixes; "all" was
treated as "process all prefixes."

After: nargs="?" with no choices — any string is accepted silently,
and "all" is now treated as a literal prefix name (will process a
non-existent all/ prefix). Old callers passing "all" will silently do the
wrong thing. At minimum, add a deprecation note or handle "all" explicitly.

5. `detect_prefixes_from_bucket` uses `BUCKET_NAME` global

paginator.paginate(Bucket=BUCKET_NAME, ...)

Fine for CLI (because initialize_bucket() runs first in main()), but for
Lambda use it depends on call order. Given that lambda_function.py is
missing, this call order is unverified.

Minor Issues

6. `update_pep503_index` docstring is a comment, not a docstring

def update_pep503_index(prefix: str, compute_sha256: bool = False, upload: bool = True):
    # Regenerates the PEP 503 simple index...

Should use """...""" for consistency with other functions in the file.

7. No tests for new auto-detection path

resolve_prefixes() and detect_prefixes_from_bucket() have no unit or
integration test coverage.

8. Blank line removed between `PREFIXES` and `CUSTOM_PREFIX`

Cosmetic, inconsistent with surrounding style.

Summary

Area	Status
Deliverable	`lambda_function.py` missing — core goal unmet
Regression	SHA256 checksums dropped from uploaded indexes
Stability	Uninitialized globals will crash non-CLI code paths
Compatibility	`all` prefix and choices validation silently removed

The update_pep503_index() extraction and initialize_bucket() refactor are
in the right direction for Lambda support, but the PR needs: the Lambda file
itself, the with_metadata fix, module-level global declarations, and
handling of the all prefix case before it's mergeable.

araravik-psd · 2026-03-25T17:44:20Z

Some additional comments from my agent:

Critical Issues

1. Missing lambda_function.py — the PR's primary goal is absent

The PR description says "Introduced a thin lambda_function.py wrapper to handle S3 events" but the diff only modifies manage.py. The Lambda integration layer never ships. Without it, this PR is a pure CLI refactor with no Lambda functionality delivered.

2. with_metadata behavior regression — SHA256 checksums dropped from normal uploads

Before:
idx = S3Index.from_S3(prefix=prefix, with_metadata=True or args.compute_sha256)
# Note: `True or X` always evaluates True — metadata always fetched
After (update_pep503_index):
idx = S3Index.from_S3(prefix=prefix, with_metadata=compute_sha256)
Metadata (SHA256 checksums) is now only fetched when --compute-sha256 is passed. For normal upload runs, compute_sha256=False, so with_metadata=False, meaning uploaded index.html files will have no #sha256=... fragments. This silently breaks PEP 503 checksum links in the package index.

The original True or args.compute_sha256 was itself a bug (always True), but the fix should be with_metadata=True for normal upload paths, not with_metadata=compute_sha256.

3. Uninitialized globals — Lambda invocations will crash with NameError

The original module-level assignments:
BUCKET_NAME = getenv("S3_BUCKET_PY", "therock-dev-python")
BUCKET = S3.Bucket(BUCKET_NAME)
INDEX_BUCKETS = {BUCKET}
...are removed and replaced with initialize_bucket(), which is only called from main(). But BUCKET_NAME, BUCKET, and INDEX_BUCKETS are no longer declared at module scope at all. Any code path (Lambda handler, import-time side effects) that uses these before initialize_bucket() runs will get NameError: name 'BUCKET_NAME' is not defined. The globals should at least be declared as None at module level.

Moderate Issues

4. Loss of prefix validation / all behavior

Before: choices=PREFIXES + ["all"] enforced known prefixes; "all" was treated as "process all prefixes."

After: nargs="?" with no choices — any string is accepted silently, and "all" is now treated as a literal prefix name (will process a non-existent all/ prefix). Old callers passing "all" will silently do the wrong thing. At minimum, add a deprecation note or handle "all" explicitly.

5. detect_prefixes_from_bucket uses BUCKET_NAME global
paginator.paginate(Bucket=BUCKET_NAME, ...)
Fine for CLI (because initialize_bucket() runs first in main()), but for Lambda use it depends on call order. Given that lambda_function.py is missing, this call order is unverified.

Minor Issues

6. update_pep503_index docstring is a comment, not a docstring
def update_pep503_index(prefix: str, compute_sha256: bool = False, upload: bool = True):
    # Regenerates the PEP 503 simple index...
Should use """...""" for consistency with other functions in the file.

7. No tests for new auto-detection path

resolve_prefixes() and detect_prefixes_from_bucket() have no unit or integration test coverage.

8. Blank line removed between PREFIXES and CUSTOM_PREFIX

Cosmetic, inconsistent with surrounding style.

Summary

Area Status
Deliverable lambda_function.py missing — core goal unmet
Regression SHA256 checksums dropped from uploaded indexes
Stability Uninitialized globals will crash non-CLI code paths
Compatibility all prefix and choices validation silently removed
The update_pep503_index() extraction and initialize_bucket() refactor are in the right direction for Lambda support, but the PR needs: the Lambda file itself, the with_metadata fix, module-level global declarations, and handling of the all prefix case before it's mergeable.

@marbre Thanks a lot for pointing these out.

I’ve already pushed updates addressing them:

Switched back to with_metadata=True so SHA256 fragments are preserved during normal uploads.
Reintroduced safe module-level initialization for BUCKET_NAME, BUCKET, and INDEX_BUCKETS to avoid any NameError before initialize_bucket() runs.
Added back explicit handling for "all" to keep previous behavior intact.
Converted the inline comment into a proper docstring for consistency.
Fixed a small bug the S3 class and added details for the same in description

For the auto-detection logic, I am running them locally, will add the outputs as a git gist of CLI test output for both v2/ and v2-staging/ in this PR to show it working as expected. I can add tests for these as a subsequent PR if that works

All of this is already committed in the latest revision.

marbre

The PR description is still misleading:

Introduced a thin lambda_function.py wrapper to handle S3 events.

This is not part of the PR, thus it shouldn't be mentioned to be part. Rather looks like a missing file. It is okay to use an agent for your PRs but you still own the code as well as any description. That said, please follow https://llvm.org/docs/AIToolPolicy.html#llvm-ai-tool-use-policy:

it is strongly recommended that contributors write PR descriptions themselves

marbre · 2026-03-26T14:32:27Z

+"""
+NOTE:
+This function is used both by the CLI (via main()) and by the AWS Lambda
+wrapper (lambda_function.py). It must remain callable without relying on
+CLI argument parsing or main() initialization.
+"""


Suggested change

"""

NOTE:

This function is used both by the CLI (via main()) and by the AWS Lambda

wrapper (lambda_function.py). It must remain callable without relying on

CLI argument parsing or main() initialization.

"""

# NOTE:

# This function is used both by the CLI (via main()) and by the AWS Lambda

# wrapper (lambda_function.py). It must remain callable without relying on

# CLI argument parsing or main() initialization.

Should not be a string.

Could also used in the function string, wdyt?

marbre · 2026-03-26T14:33:13Z

+# NOTE:
+# initialize_bucket() is required for both CLI and Lambda usage.
+# Do not move bucket initialization exclusively into main(), as the
+# Lambda wrapper imports and calls update_pep503_index() directly.


Why not move this into the function description (currently there is none).

I think you forgot to delete this after added into the function. I see the same sentence in docstring.

marbre · 2026-03-26T14:33:50Z

+        "Base prefix for auto-detection (e.g. v2/, v2-staging/, v3/whl/). "
+        "Required when using --auto-detect-prefixes."


Missing indention:

Suggested change

"Base prefix for auto-detection (e.g. v2/, v2-staging/, v3/whl/). "

"Required when using --auto-detect-prefixes."

"Base prefix for auto-detection (e.g. v2/, v2-staging/, v3/whl/). "

"Required when using --auto-detect-prefixes."

marbre · 2026-04-01T12:32:55Z

@erman-gurses @HereThereBeDragons @amd-shiraz please review before I do a final round.

erman-gurses

Added some comments - they are minor issues in general.

erman-gurses · 2026-04-02T22:49:25Z

+    initialize_bucket() is required for both CLI and Lambda usage.
+    Do not move bucket initialization exclusively into main(), as the
+    Lambda wrapper imports and calls update_pep503_index() directly.
+   '''


Nit: indentation slightly off for '''

erman-gurses · 2026-04-02T22:55:38Z

+# Must be provided via --bucket (CLI) or S3_BUCKET_PY (env).
+
+# Bucket globals configured via initialize_bucket()
+BUCKET_NAME: Optional[str] = None


Based on style guide, we should not use Optional https://github.com/ROCm/TheRock/blob/main/docs/development/style_guides/python_style_guide.md#add-specific-type-hints-liberally

Updated the new annotations to use T | None per the style guide. Optional is still used elsewhere in the file, so I’ve kept the import to avoid unrelated refactors.

erman-gurses · 2026-04-02T22:57:06Z

+# NOTE:
+# initialize_bucket() is required for both CLI and Lambda usage.
+# Do not move bucket initialization exclusively into main(), as the
+# Lambda wrapper imports and calls update_pep503_index() directly.


I think you forgot to delete this after added into the function. I see the same sentence in docstring.

HereThereBeDragons

only some minor improvments. and please also have a look at erman s comments.

amd-shiraz · 2026-04-28T04:42:44Z

@araravik-psd is this PR ready to land after approval ? If yes can do another pass.

erman-gurses

LGTM, only some minor issues and questions.
Also, did we test it fully after the new changes applied?

erman-gurses · 2026-04-28T06:42:20Z

        for obj in all_sorted_packages:
            full_package_name = path.basename(obj)
            package_name = full_package_name.split('-')[0]
+            print(f"[DEBUG] Evaluating package: {package_name}")


Is this temporary? If not, we might remove it.

Yes, this should be removed

Suggested change

print(f"[DEBUG] Evaluating package: {package_name}")

marbre · 2026-04-28T14:03:19Z

+# Allow new ROCm multi-arch + device packages
+ROCM_PACKAGE_PREFIXES = (
+    "rocm_",
+    "rocm-sdk",          # safety for normalized names
+    "rocm_sdk_device",
+    "amd_torch_device",
+)


What's the value of having this in an extra list? Some of those are already in PACKAGE_ALLOW_LIST and amd_torch_device is actually not a ROCm package.

We further need to add amd_torchvision_device but the place here seems wrong.

Thanks @marbre Removed the extra list and extended the existing allow-list logic instead. Added explicit handling for amd_torchvision_device* and similar dynamic packages.

HereThereBeDragons · 2026-04-28T17:21:45Z

+            # Allow legacy packages + explicitly handle dynamic multi-arch packages
+            if (
+                pkg in PACKAGE_ALLOW_LIST
+                or pkg.startswith("rocm_sdk_device_")


why not add those to PACKAGE_ALLOW_LIST?

These are dynamic arch-suffixed packages (_gfx), so adding them to PACKAGE_ALLOW_LIST would require enumerating all variants. Using startswith avoids needing updates for every new architecture. This was my thinking here

If this is the case how have the rocm-sdk-libraries-gfx* packages ever worked?

Good catch, looking at the actual wheel names again, the single-arch rocm_sdk_libraries_gfx* packages are also dynamically suffixed similar to the rocm_sdk_device_* wheels.

I am testing this now without the startswith() handling to verify whether the existing normalization/filtering logic already covers these correctly before keeping the additional prefix checks.

@marbre @HereThereBeDragons I re-checked the original logic and rocm_sdk_libraries_gfx* packages were already being special-cased via:

match(r"rocm_sdk_libraries_gfx", package_name.lower())
https://github.com/ROCm/TheRock/blob/main/build_tools/third_party/s3_management/manage.py#L216

So they were never handled through PACKAGE_ALLOW_LIST.

The new startswith() handling is effectively extending the same pattern to the newer dynamically suffixed multi-arch package families like rocm_sdk_device_* and amd_torch_device_*.

This reverts commit c784a0f.

marbre

The currently deployed Lambda creates duplicates:

This was fixed with commit 319acc7 but the branch was never updated. I rebased and force pushed as we need to re-deploy the updated Lambda asap.

marbre

There are still a few minor issues to address. @araravik-psd please follow up and let's land this latest tomorrow.

marbre · 2026-05-07T11:35:40Z

        for obj in all_sorted_packages:
            full_package_name = path.basename(obj)
            package_name = full_package_name.split('-')[0]
+            print(f"[DEBUG] Evaluating package: {package_name}")


Yes, this should be removed

Suggested change

print(f"[DEBUG] Evaluating package: {package_name}")

marbre · 2026-05-07T11:36:15Z

+            # Allow legacy packages + explicitly handle dynamic multi-arch packages
+            if (
+                pkg in PACKAGE_ALLOW_LIST
+                or pkg.startswith("rocm_sdk_device_")


If this is the case how have the rocm-sdk-libraries-gfx* packages ever worked?

aravind-ravi1206 added 3 commits March 3, 2026 23:04

manage.py changes for lambda compatibility

4694f8e

Restore prefixes

86e375f

Update comments

1b0f3fc

github-project-automation Bot added this to TheRock Triage Mar 3, 2026

github-project-automation Bot moved this to TODO in TheRock Triage Mar 3, 2026

araravik-psd requested review from HereThereBeDragons, amd-shiraz and marbre March 3, 2026 23:17

HereThereBeDragons reviewed Mar 4, 2026

View reviewed changes

Addressing PR comments

65389fb

araravik-psd requested a review from HereThereBeDragons March 4, 2026 16:59

marbre reviewed Mar 9, 2026

View reviewed changes

Implementing PR comments

d71394a

araravik-psd requested a review from marbre March 16, 2026 21:15

Add comments for bucket name resolution

72f1cc9

marbre reviewed Mar 25, 2026

View reviewed changes

araravik-psd changed the title ~~Move S3 Python index generation to event-driven Lambda with thin wrapper architecture~~ Enable event-driven Lambda execution for S3 Python index generation Mar 25, 2026

Addressed PR comments and fixes

17fcf8a

araravik-psd requested a review from erman-gurses March 25, 2026 21:12

Reverting S3 Class changed

4dff38c

araravik-psd requested a review from marbre March 25, 2026 23:53

marbre reviewed Mar 26, 2026

View reviewed changes

Addressed PR comments

5bb2e19

araravik-psd requested a review from marbre March 30, 2026 14:36

erman-gurses reviewed Apr 2, 2026

View reviewed changes

HereThereBeDragons reviewed Apr 7, 2026

View reviewed changes

Addressing PR comments

6e2f455

araravik-psd requested review from HereThereBeDragons and erman-gurses April 15, 2026 14:29

Update Package selection for multi-arch

d8d8972

erman-gurses approved these changes Apr 28, 2026

View reviewed changes

marbre reviewed Apr 28, 2026

View reviewed changes

Addressing PR comments for ROCM_PACKAGE_PREFIXES

1126ab6

araravik-psd requested a review from marbre April 28, 2026 16:11

Adding amd_apex_device packages

c784a0f

araravik-psd mentioned this pull request Apr 28, 2026

Remove Python index generation(manage.py) from CI as it is moved over to S3-triggered Lambda #4900

Open

HereThereBeDragons reviewed Apr 28, 2026

View reviewed changes

Revert "Adding amd_apex_device packages"

445bb8c

This reverts commit c784a0f.

marbre force-pushed the users/arravikum/manage_lambda_changes branch from 445bb8c to c4f34fe Compare April 29, 2026 19:43

marbre reviewed Apr 29, 2026

View reviewed changes

marbre reviewed May 7, 2026

View reviewed changes

Addressed PR comments and added more archs

2360e78

araravik-psd force-pushed the users/arravikum/manage_lambda_changes branch from 5471a43 to 2360e78 Compare May 7, 2026 15:11

		# TODO: bucket mirror just to hold index used with CDN
		# BUCKET_CDN = S3.Bucket('therock-nightly-python-testing')

		"Base prefix for auto-detection (e.g. v2/, v2-staging/, v3/whl/). "
		"Required when using --auto-detect-prefixes."

Conversation

araravik-psd commented Mar 3, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

araravik-psd commented Mar 3, 2026

Uh oh!

amd-shiraz commented Mar 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

HereThereBeDragons left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

marbre Mar 4, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Uh oh!

marbre left a comment

Choose a reason for hiding this comment

Critical Issues

1. Missing lambda_function.py — the PR's primary goal is absent

2. with_metadata behavior regression — SHA256 checksums dropped from normal uploads

3. Uninitialized globals — Lambda invocations will crash with NameError

Moderate Issues

4. Loss of prefix validation / all behavior

5. detect_prefixes_from_bucket uses BUCKET_NAME global

Minor Issues

6. update_pep503_index docstring is a comment, not a docstring

7. No tests for new auto-detection path

8. Blank line removed between PREFIXES and CUSTOM_PREFIX

Summary

Uh oh!

araravik-psd commented Mar 25, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Critical Issues

1. Missing lambda_function.py — the PR's primary goal is absent

2. with_metadata behavior regression — SHA256 checksums dropped from normal uploads

3. Uninitialized globals — Lambda invocations will crash with NameError

Moderate Issues

4. Loss of prefix validation / all behavior

5. detect_prefixes_from_bucket uses BUCKET_NAME global

Minor Issues

6. update_pep503_index docstring is a comment, not a docstring

7. No tests for new auto-detection path

8. Blank line removed between PREFIXES and CUSTOM_PREFIX

Summary

Uh oh!

marbre left a comment

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Choose a reason for hiding this comment

Uh oh!

marbre commented Apr 1, 2026

Uh oh!

erman-gurses left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Choose a reason for hiding this comment

araravik-psd commented Mar 3, 2026 •

edited

Loading

amd-shiraz commented Mar 4, 2026 •

edited

Loading

marbre Mar 4, 2026 •

edited

Loading

1. Missing `lambda_function.py` — the PR's primary goal is absent

2. `with_metadata` behavior regression — SHA256 checksums dropped from normal uploads

3. Uninitialized globals — Lambda invocations will crash with `NameError`

4. Loss of `prefix` validation / `all` behavior

5. `detect_prefixes_from_bucket` uses `BUCKET_NAME` global

6. `update_pep503_index` docstring is a comment, not a docstring

8. Blank line removed between `PREFIXES` and `CUSTOM_PREFIX`

araravik-psd commented Mar 25, 2026 •

edited

Loading

1. Missing `lambda_function.py` — the PR's primary goal is absent

2. `with_metadata` behavior regression — SHA256 checksums dropped from normal uploads

3. Uninitialized globals — Lambda invocations will crash with `NameError`

4. Loss of `prefix` validation / `all` behavior

5. `detect_prefixes_from_bucket` uses `BUCKET_NAME` global

6. `update_pep503_index` docstring is a comment, not a docstring

8. Blank line removed between `PREFIXES` and `CUSTOM_PREFIX`

erman-gurses left a comment •

edited

Loading

erman-gurses left a comment •

edited

Loading