PEP458: Update TUF repository metadata on project index change #15815

lukpueh · 2024-04-18T13:30:54Z

EDIT 2024/06/21: update description; drop discussion points (see #15815 (comment), #15815 (comment))

Part 3 in a series of PRs to integrate Repository Service for TUF (RSTUF) with Warehouse for PEP 458 adoption. Previous PR was #15484

This PR adds an async task to the tuf subpackage, in order to capture project changes in TUF repository metadata via the RSTUF API. Relevant Warehouse views -- file upload and project management (remove, yank) -- are updated to trigger the task.

This PR also updates the so far unused render_simple_detail function for this purpose.

simi · 2024-04-25T11:04:20Z

dev/environment

@@ -64,6 +64,9 @@ TWOFACTORMANDATE_AVAILABLE=true
 TWOFACTORMANDATE_ENABLED=true
 OIDC_AUDIENCE=pypi

+TUF_RSTUF_API_URL="http://rstuf-api"
+TUF_ENABLED=true


Btw. in RubyGems.org we use RSTUF_API_URL and we enable it just by presence. Not sure if worth to keep it similar.

https://github.com/rubygems/rubygems.org/blob/ad1269b77e3ac83edac6a8f399125ba9fec0280d/config/initializers/rstuf.rb#L3C9-L7

simi · 2024-04-25T11:06:49Z

warehouse/tuf/__init__.py

+
+    Raises RuntimeError, if task fails.
+    """
+    while True:


Is this in theory of state not being changed infinite loop?

Yeah, this is just a stopgap solution. We should definitely do better. See discussion item about async task handling in PR description.

lukpueh · 2024-05-02T15:27:37Z

Discussion

There are a few open question, for which I'd appreciate input from folks, who are more familiar with Warehouse.

...

Any chance to get some feedback from Warehouse maintainers? Also happy to arrange a call to discuss in person. cc @ewdurbin

On a side-note, we released v1.0.0 of the crypto interface used by RSTUF earlier today, and it includes a "VaultSigner", tailored for signing TUF metadata with a HashiCorp Vault key. :)

ewdurbin · 2024-06-03T11:50:57Z

Completeness
The PR should trigger TUF metadata updates in all relevant views, but someone needs to confirm. Basically, we need to update TUF metadata whenever a project change would result in a different project simple detail file, e.g. on distribution file upload, file removal, or yank. I wonder, if we could register a db hook for those changes, e.g. after_commit, which would also give us better certainty that the db change isn't rolled back.

It may be worth considering the machinery that handles our CDN cache purges (see warehouse/cache/origin) though it is admittedly opaque and a little convoluted to store "purge keys" on the request for handling after flush. This is woefully undocumented, but has been functionally reliable for a long time.

Async task handling
Metadata update tasks in RSTUF are async. This PR busy polls the RSTUF task status API until the corresponding metadata update task has finished. I'm sure, Warehouse has better ways of dealing with async tasks.

Yes, look for usage of the @tasks.task decorator throughout the codebase.

@facutuesca

Previously, `repository-service-tuf` (i.e. the RSTUF cli) was used to bootstrap an RSTUF repo for development. This PR re-implements the relevant parts of the cli locally in Warehouse and removes the `repository-service-tuf` dependency, which conflicts with other dependencies. Change details - Add lightweight RSTUF API client library (can be re-used for pypi#15815) - Add local `warehouse tuf bootstrap` cli subcommand, to wraps lib calls - Invoke local cli via `make inittuf` - Remove dependency supersedes pypi#15958 (cc @facutuesca @woodruffw) Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

@facutuesca

Previously, `repository-service-tuf` (i.e. the RSTUF cli) was used to bootstrap an RSTUF repo for development. This PR re-implements the relevant parts of the cli locally in Warehouse and removes the `repository-service-tuf` dependency, which conflicts with other dependencies. Change details - Add lightweight RSTUF API client library (can be re-used for pypi#15815) - Add local `warehouse tuf bootstrap` cli subcommand, to wraps lib calls - Invoke local cli via `make inittuf` - Remove dependency supersedes pypi#15958 (cc @facutuesca @woodruffw) Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

@facutuesca

* Replace conflicting repository-service-tuf dep Previously, `repository-service-tuf` (i.e. the RSTUF cli) was used to bootstrap an RSTUF repo for development. This PR re-implements the relevant parts of the cli locally in Warehouse and removes the `repository-service-tuf` dependency, which conflicts with other dependencies. Change details - Add lightweight RSTUF API client library (can be re-used for #15815) - Add local `warehouse tuf bootstrap` cli subcommand, to wraps lib calls - Invoke local cli via `make inittuf` - Remove dependency supersedes #15958 (cc @facutuesca @woodruffw) Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu> * Make payload arg in tuf cli "lazy" Other than the regular click File, the LazyFile also has the "name" attribute, when passing stdin via "-". We print the name on success. Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu> * Add minimal unittest for TUF bootstrap cli Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu> * Add unit tests for RSTUF API client lib Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu> --------- Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

lukpueh · 2024-06-20T17:22:03Z

Just rebased on latest main, resolving conflicts and including some minor refactorings in the original commits for consistency with #16098. More significant changes/additions are added on top. The commit history is now a bit evolutionary. Happy to rewrite/squash, if that's a requirement.

cc @ofek, @trishankatdatadog, @dstufft

lukpueh · 2024-06-21T09:54:28Z

Completeness
The PR should trigger TUF metadata updates in all relevant views, but someone needs to confirm. Basically, we need to update TUF metadata whenever a project change would result in a different project simple detail file, e.g. on distribution file upload, file removal, or yank. I wonder, if we could register a db hook for those changes, e.g. after_commit, which would also give us better certainty that the db change isn't rolled back.

It may be worth considering the machinery that handles our CDN cache purges (see warehouse/cache/origin) though it is admittedly opaque and a little convoluted to store "purge keys" on the request for handling after flush. This is woefully undocumented, but has been functionally reliable for a long time.

Thanks for the pointer, @ewdurbin! I decided to leave the explicit invocations for now. We can improve this in a follow-up PR.

Async task handling
Metadata update tasks in RSTUF are async. This PR busy polls the RSTUF task status API until the corresponding metadata update task has finished. I'm sure, Warehouse has better ways of dealing with async tasks.

Yes, look for usage of the @tasks.task decorator throughout the codebase.

Thanks, this worked nicely.

lukpueh · 2024-06-21T09:58:26Z

Error handling
This means setting up proper retry mechanism for temporary errors, and figuring out user feedback and recovery for permanent errors.

Now that the RSTUF API is called in an async task errors shouldn't affect the normal operation of Warehouse. I suggest to take the PR as is and improve error handling (ignore, retry or notify) in a next iteration.

lukpueh · 2024-06-21T10:16:48Z

(sorry for the noise, I accidentally hit the "close with comment" button)

trishankatdatadog · 2024-07-02T21:45:11Z

Just rebased on latest main, resolving conflicts and including some minor refactorings in the original commits for consistency with #16098. More significant changes/additions are added on top. The commit history is now a bit evolutionary. Happy to rewrite/squash, if that's a requirement.

cc @ofek, @trishankatdatadog, @dstufft

Sorry, was on PTO. How can we best help here? For example, would you like us (esp. Donald) to review?

lukpueh · 2024-07-03T08:37:55Z

Thanks, @trishankatdatadog! A review by anyone would be amazing. Donald seems to be a great candidate because he knows Warehouse and TUF extremely well.

As said above, I think, this PR can be merged as is. However, I'd appreciate help with improvements in follow-up PRs (see my comments above about completeness and error handling).

trishankatdatadog

LGTM FWIW!

My high-level Q is: does this work with the JSON Simple API (particularly for PEP 740)?

trishankatdatadog · 2024-07-29T19:39:16Z

warehouse/manage/views/__init__.py

@@ -1861,6 +1862,8 @@ def delete_project(project, request):

    remove_project(project, request)

+    request.task(tuf.update_metadata).delay(project.id)


I see these calls a fair bit. Is there anything we could automatically do to reduce the burden of the user?

See discussion about completeness above: #15815 (comment)

TLDR: This should be improved, but could happen in a follow-up PR.

I'll echo this - we don't have many tasks directly called inside views today, rather we listen to the objects being changed via sqlalchemy and trigger tasks that way.

Here's a couple of examples:

warehouse/warehouse/search/__init__.py

Lines 27 to 64 in e8195e4

@db.listens_for(db.Session, "after_flush")

def store_projects_for_project_reindex(config, session, flush_context):

# We'll (ab)use the session.info dictionary to store a list of pending

# Project updates to the session.

projects_to_update = session.info.setdefault(

"warehouse.search.project_updates", set()

)

projects_to_delete = session.info.setdefault(

"warehouse.search.project_deletes", set()

)

# Go through each new, changed, and deleted object and attempt to store

# a Project to reindex for when the session has been committed.

for obj in session.new | session.dirty:

if obj.__class__ == Project:

projects_to_update.add(obj)

if obj.__class__ == Release:

projects_to_update.add(obj.project)

for obj in session.deleted:

if obj.__class__ == Project:

projects_to_delete.add(obj)

if obj.__class__ == Release:

projects_to_update.add(obj.project)

@db.listens_for(db.Session, "after_commit")

def execute_project_reindex(config, session):

projects_to_update = session.info.pop("warehouse.search.project_updates", set())

projects_to_delete = session.info.pop("warehouse.search.project_deletes", set())

from warehouse.search.tasks import reindex_project, unindex_project

for project in projects_to_update:

config.task(reindex_project).delay(project.normalized_name)

for project in projects_to_delete:

config.task(unindex_project).delay(project.normalized_name)

warehouse/warehouse/observations/tasks.py

Lines 33 to 54 in e8195e4

@db.listens_for(db.Session, "after_flush")

def new_observation_created(_config, session: SA_Session, _flush_context):

# Go through each new, changed, and deleted object and attempt to store

# a cache key that we'll want to purge when the session has been committed.

for obj in session.new:

if isinstance(obj, Observation):

# Add to `session.info` so we can access it in the after_commit listener.

session.info.setdefault("warehouse.observations.new", set()).add(obj)

@db.listens_for(db.Session, "after_commit")

def execute_observation_report(config: Configurator, session: SA_Session):

# Fetch the observations from the session.

observations = session.info.pop("warehouse.observations.new", set())

for obj in observations:

# We pass the ID of the Observation, not the Observation itself,

# because the Observation object is not currently JSON-serializable.

config.task(report_observation_to_helpscout).delay(obj.id)

@tasks.task(bind=True, ignore_result=True, acks_late=True)

def report_observation_to_helpscout(task, request: Request, model_id: UUID) -> None:

trishankatdatadog · 2024-07-29T19:41:00Z

tests/unit/packaging/test_utils.py

@@ -90,7 +91,7 @@ def test_render_simple_detail_with_store(db_request, monkeypatch, jinja):

    fake_named_temporary_file = pretend.stub(
        name="/tmp/wutang",
-        write=pretend.call_recorder(lambda data: None),
+        write=pretend.call_recorder(lambda data: 42),


Nitpick: could we specify and reuse a constant for the size (42)?

lukpueh · 2024-07-30T08:32:44Z

Thanks for the review, @trishankatdatadog!

My high-level Q is: does this work with the JSON Simple API (particularly for PEP 740)?

Great catch! This PR only covers the html simple index, but could be easily extended to also include the json simple index.

AFAICS we'd only have to make sure that

project json simple index is persistent (for some time) like the html simple index is, and
post two TUF targets instead of one.

PEP 691 that TUF uses these target paths:

simple/PROJECT/vnd.pypi.simple.v1.html
simple/PROJECT/vnd.pypi.simple.v1.json

trishankatdatadog · 2024-07-30T12:21:06Z

AFAICS we'd only have to make sure that

Got it, thanks. @woodruffw is this correct: did we miss anything?

miketheman

Hey gang! I reviewed, and commented on how this is constructed, let me know if anything isn't clear!

miketheman · 2024-08-23T20:34:32Z

dev/environment

@@ -64,6 +64,8 @@ TWOFACTORMANDATE_AVAILABLE=true
 TWOFACTORMANDATE_ENABLED=true
 OIDC_AUDIENCE=pypi

+TUF_RSTUF_API_URL="http://rstuf-api"


question: curious why TUF is prefixing RSTUF? Shouldn't this be RSTUF_API_URL instead?

Solved with commit refactor: remove prefixes tuf/TUF from config

miketheman · 2024-08-23T20:40:27Z

warehouse/tuf/__init__.py

This init file is now starting to add more concepts other than the only pieces that are necessary to bootstrap.

I'd like to see a tuf/services.py and a tuf/interfaces.py, similar to other services, and split the components that way. We use pyramid-services style for dependency injection to make remote services either available or unavailable during dev/test - see examples in email or captcha modules.

Then, anything specific to the TUF CLI can get moved to the right module, and keep __init__.py for things that are only needed on import.

Done with commit chore: Implement TUF Interface and RSTUF Service

miketheman · 2024-08-23T20:42:46Z

warehouse/tuf/__init__.py

+@tasks.task(ignore_result=True, acks_late=True)
+def update_metadata(request: Request, project_id: UUID):


Beyond services and interfaces, split out another module for tasks.py to keep the modules scoped

Solved with commit refactor: split tuf.__init__.tasks to tuf.tasks

miketheman · 2024-08-23T21:53:44Z

warehouse/config.py

@@ -367,6 +367,7 @@ def configure(settings=None):
        "OIDC_BACKEND",
        default="warehouse.oidc.services.OIDCPublisherService",
    )
+    maybe_set(settings, "tuf.rstuf_api_url", "TUF_RSTUF_API_URL")


Similar comment on naming. This might be better expressed as rstuf.api_url, depending on which direction we go.

Solved with commit refactor: remove prefixes tuf/TUF from config

miketheman · 2024-08-23T21:54:58Z

warehouse/forklift/legacy.py

@@ -1168,6 +1169,8 @@ def file_upload(request):

    request.db.flush()  # flush db now so server default values are populated for celery

+    request.task(tuf.update_metadata).delay(release.project.id)


This should probably be wrapped in a conditional that does not create the task if the URL setting is not created so file uploads are not creating tasks that get disposed immediately.

Solved with commit chore: wrapped metadata update if rstuf is enabled

miketheman · 2024-08-23T21:57:19Z

tests/unit/tuf/test_tuf.py

+        # Test fail with incomplete response json (i.e. no bootstrap error)
+        del resp_json["data"]
+        with pytest.raises(tuf.RSTUFNoBootstrapError):
+            tuf.post_artifacts(self.server, payload)


Extract failure test conditions to their own test cases

Hi @miketheman,

It has changed with my refactoring, but should I create two separate tests?

warehouse/tests/unit/tuf/test_services.py

Lines 127 to 160 in 86803ac

@pytest.mark.parametrize(

("states", "exception", "message"),

[

(

[

{"data": {"state": "PENDING"}},

{"data": {"state": "STARTED"}},

{"data": {"state": "RECEIVED"}},

{"data": {"state": "STARTED"}},

{"data": {"state": "SUCCESS"}},

],

None,

"",

),

(

[

{"data": {"state": "PENDING"}},

{"data": {"state": "STARTED"}},

{"data": {"state": "RECEIVED"}},

{"data": {"state": "STARTED"}},

{"data": {"state": "FAILURE"}},

],

services.RSTUFError,

"RSTUF job failed, please check payload and retry",

),

(

[

{"data": {"state": "PENDING"}},

{"data": {"state": "STARTED"}},

{"data": {"state": "RECEIVED"}},

{"data": {"state": "STARTED"}},

{"data": {"state": "ERRORED"}},

],

services.RSTUFError,

miketheman · 2024-08-23T21:58:18Z

tests/unit/tuf/test_tuf.py

+        # Test early return, if no RSTUF API URL configured
+        db_request.registry.settings = {"tuf.rstuf_api_url": None}
+        tuf.update_metadata(db_request, project_id)
+
+        assert not one.calls


Another candidate for splitting to another test case

Solved with commit refactor: split tuf.__init__.tasks to tuf.tasks

miketheman · 2024-08-23T22:01:07Z

warehouse/tuf/__init__.py

+    # NOTE: We ignore the returned simple detail path with the content hash as
+    # infix. In TUF metadata the project name and hash are listed separately, so
+    # that there is only one entry per target file, even if the content changes.
+    digest, _, size = render_simple_detail(project, request, store=True)


If we explicitly ignore the simple_detail_path, why should the function continue to return it at all? Does any other caller use it?

Removed in the commit refactor: utils.render_simple_detail

miketheman · 2024-08-23T22:12:45Z

warehouse/packaging/utils.py

+    with tempfile.NamedTemporaryFile() as f:
+        simple_detail_size = f.write(content.encode("utf-8"))
+        f.flush()


Is this the only way for us to measure the size by writing out to disk? Could we keep the writing to disk still behind the if store conditional, and calculate the size in memory?

Addressed in refactor: utils.render_simple_detail

miketheman · 2024-08-23T22:21:56Z

warehouse/manage/views/__init__.py

@@ -1861,6 +1862,8 @@ def delete_project(project, request):

    remove_project(project, request)

+    request.task(tuf.update_metadata).delay(project.id)


I'll echo this - we don't have many tasks directly called inside views today, rather we listen to the objects being changed via sqlalchemy and trigger tasks that way.

Here's a couple of examples:

warehouse/warehouse/search/__init__.py

Lines 27 to 64 in e8195e4

@db.listens_for(db.Session, "after_flush")

def store_projects_for_project_reindex(config, session, flush_context):

# We'll (ab)use the session.info dictionary to store a list of pending

# Project updates to the session.

projects_to_update = session.info.setdefault(

"warehouse.search.project_updates", set()

)

projects_to_delete = session.info.setdefault(

"warehouse.search.project_deletes", set()

)

# Go through each new, changed, and deleted object and attempt to store

# a Project to reindex for when the session has been committed.

for obj in session.new | session.dirty:

if obj.__class__ == Project:

projects_to_update.add(obj)

if obj.__class__ == Release:

projects_to_update.add(obj.project)

for obj in session.deleted:

if obj.__class__ == Project:

projects_to_delete.add(obj)

if obj.__class__ == Release:

projects_to_update.add(obj.project)

@db.listens_for(db.Session, "after_commit")

def execute_project_reindex(config, session):

projects_to_update = session.info.pop("warehouse.search.project_updates", set())

projects_to_delete = session.info.pop("warehouse.search.project_deletes", set())

from warehouse.search.tasks import reindex_project, unindex_project

for project in projects_to_update:

config.task(reindex_project).delay(project.normalized_name)

for project in projects_to_delete:

config.task(unindex_project).delay(project.normalized_name)

warehouse/warehouse/observations/tasks.py

Lines 33 to 54 in e8195e4

@db.listens_for(db.Session, "after_flush")

def new_observation_created(_config, session: SA_Session, _flush_context):

# Go through each new, changed, and deleted object and attempt to store

# a cache key that we'll want to purge when the session has been committed.

for obj in session.new:

if isinstance(obj, Observation):

# Add to `session.info` so we can access it in the after_commit listener.

session.info.setdefault("warehouse.observations.new", set()).add(obj)

@db.listens_for(db.Session, "after_commit")

def execute_observation_report(config: Configurator, session: SA_Session):

# Fetch the observations from the session.

observations = session.info.pop("warehouse.observations.new", set())

for obj in observations:

# We pass the ID of the Observation, not the Observation itself,

# because the Observation object is not currently JSON-serializable.

config.task(report_observation_to_helpscout).delay(obj.id)

@tasks.task(bind=True, ignore_result=True, acks_late=True)

def report_observation_to_helpscout(task, request: Request, model_id: UUID) -> None:

miketheman

I can't seem to use GitHub correctly, gonna stop for today 😆

because I am making mistakes

lukpueh · 2024-08-28T13:42:41Z

Thanks for the feedback, @miketheman!

@kairoaraujo, @trishankatdatadog, and friends, please feel free to take over as discussed the other day!

kairoaraujo · 2024-10-04T10:26:49Z

Note: I pushed some comment fixes, but I am still not finished (Work in Progress)

The simple detail file size is to be included in TUF metadata, along with its path and hash. Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Add new TUF subpackage with functions to talk to the RSTUF API, in order to update TUF repository metadata when a project changes. The function requires RSTUF to be bootstrapped (see `make inittuf`), and can additionally be toggled via setting: - disabled by default - enabled in dev environment Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Relevant views are "file upload", "remove" (file, release, project), "yank". Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Make `tuf.update_metadata` an async task and call with `delay`. The `project` argument becomes `project_id`, because it must be serializable. This change allows us to remove the `disable_tuf` test fixture, because `delay` is mocked in the relevant tests anyway, so that `update_metadata` isn't actually called in any existing tests. However, call_recorder asserts need to be and are adapted. Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Instead, allow handling a custom RSTUFNoBootstrapError, which is raised, if tuf.update_metadata is called before RSTUF is bootstrapped. Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

@simi

Using TUF_RSTUF_API_URL as toggle instead should be good enough. This is also what rubygems does. h/t @simi Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

@trishankatdatadog

requested by @trishankatdatadog Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

- `tuf.rstuf_*` to `rstuf.*` - environment variable from `TUF_RSTUF*` to `RSTUF*` Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

this refactor split the tasks from the tuf.__init__.py to tasks.py Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

use rstuf upstream cli update `make inittuf` to use upstream command Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

- Implements the tuf/interfaces.py with ITUFService - Implements the tuf/services.py with RSTUFService - Refactor the tuf/tasks.py metadata_update to use the Service - RSTUF tasks are triggered by changes in the Project db objects - Add unit tests Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

- remove the simple_detail_path as it is not used by any other caller - calculate size without writting the file, only write when store is true Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

- Update RSTUF services to 1.0.0b1 - Update payload to use `artifacts` instead `targets` - Add RSTUF `PRE_RUN` state Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

kairoaraujo · 2024-10-09T10:47:04Z

Hey folks,
Ready for another review! 🎉

woodruffw · 2024-10-10T08:18:19Z

warehouse/tuf/tasks.py

+    # NOTE: We ignore the returned simple detail path with the content hash as
+    # infix. In TUF metadata the project name and hash are listed separately, so
+    # that there is only one entry per target file, even if the content changes.
+    digest, size = render_simple_detail(project, request, store=True)


How does this interact with/handle the other artifacts/pieces of metadata produced on file upload? In particular, I'm thinking of:

The JSON (PEP 691) simple index

The detached metadata (PEP 658)

lukpueh force-pushed the rstuf-add-targets branch from ede9c1a to 33dabfe Compare April 22, 2024 14:05

simi reviewed Apr 25, 2024

View reviewed changes

lukpueh mentioned this pull request Jun 13, 2024

Replace conflicting repository-service-tuf dep #16098

Merged

lukpueh force-pushed the rstuf-add-targets branch from 33dabfe to 34a749f Compare June 20, 2024 17:15

lukpueh marked this pull request as ready for review June 20, 2024 17:15

lukpueh requested a review from a team as a code owner June 20, 2024 17:15

lukpueh force-pushed the rstuf-add-targets branch from 6f89d17 to 106a536 Compare June 21, 2024 10:06

lukpueh closed this Jun 21, 2024

lukpueh reopened this Jun 21, 2024

lukpueh force-pushed the rstuf-add-targets branch from 106a536 to 6dfb07d Compare July 1, 2024 11:26

ewdurbin self-requested a review July 1, 2024 16:42

trishankatdatadog reviewed Jul 29, 2024

View reviewed changes

miketheman previously approved these changes Aug 23, 2024

View reviewed changes

This comment was marked as outdated.

Sign in to view

miketheman self-requested a review August 23, 2024 22:23

miketheman reviewed Aug 23, 2024

View reviewed changes

kairoaraujo force-pushed the rstuf-add-targets branch 2 times, most recently from c39a363 to 28ab72d Compare October 4, 2024 09:58

kairoaraujo force-pushed the rstuf-add-targets branch 2 times, most recently from bdc3309 to 0498c0a Compare October 7, 2024 10:18

lukpueh and others added 15 commits October 8, 2024 08:45

Return file size from render_simple_detail

6be8b86

The simple detail file size is to be included in TUF metadata, along with its path and hash. Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Call TUF metadata update function in relevant views

e381cff

Relevant views are "file upload", "remove" (file, release, project), "yank". Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Update TUF section in developer documentation

c982fe4

Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Do not check if RSTUF is bootstrapped

b95f1a6

Instead, allow handling a custom RSTUFNoBootstrapError, which is raised, if tuf.update_metadata is called before RSTUF is bootstrapped. Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Drop TUF_ENABLED setting

db8e2c0

Using TUF_RSTUF_API_URL as toggle instead should be good enough. This is also what rubygems does. h/t @simi Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Add tuf tests

bdbcf09

Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

make translations

08d5bd1

Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

Use variable for test value in a TUF test

31e68a9

requested by @trishankatdatadog Signed-off-by: Lukas Puehringer <lukas.puehringer@nyu.edu>

refactor: remove prefixes tuf/TUF from config

1b92870

- `tuf.rstuf_*` to `rstuf.*` - environment variable from `TUF_RSTUF*` to `RSTUF*` Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

refactor: split tuf.__init__.tasks to tuf.tasks

825cb53

this refactor split the tasks from the tuf.__init__.py to tasks.py Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

cli: remove rstuf custom command add upstream

e9183e5

use rstuf upstream cli update `make inittuf` to use upstream command Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

refactor: utils.render_simple_detail

f982336

- remove the simple_detail_path as it is not used by any other caller - calculate size without writting the file, only write when store is true Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

kairoaraujo force-pushed the rstuf-add-targets branch from 0498c0a to f982336 Compare October 9, 2024 08:01

kairoaraujo added 2 commits October 9, 2024 12:01

chore: bump RSTUF API and Worker to 1.0.0b1

43edb0a

- Update RSTUF services to 1.0.0b1 - Update payload to use `artifacts` instead `targets` - Add RSTUF `PRE_RUN` state Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

translations: make translations

a0816e2

Signed-off-by: Kairo de Araujo <kairo@dearaujo.nl>

Merge branch 'main' into rstuf-add-targets

2086b66

woodruffw reviewed Oct 10, 2024

View reviewed changes

		@@ -1861,6 +1862,8 @@ def delete_project(project, request):

		remove_project(project, request)

		request.task(tuf.update_metadata).delay(project.id)

	@db.listens_for(db.Session, "after_flush")
	def store_projects_for_project_reindex(config, session, flush_context):
	# We'll (ab)use the session.info dictionary to store a list of pending
	# Project updates to the session.
	projects_to_update = session.info.setdefault(
	"warehouse.search.project_updates", set()
	)
	projects_to_delete = session.info.setdefault(
	"warehouse.search.project_deletes", set()
	)

	# Go through each new, changed, and deleted object and attempt to store
	# a Project to reindex for when the session has been committed.
	for obj in session.new \| session.dirty:
	if obj.__class__ == Project:
	projects_to_update.add(obj)
	if obj.__class__ == Release:
	projects_to_update.add(obj.project)

	for obj in session.deleted:
	if obj.__class__ == Project:
	projects_to_delete.add(obj)
	if obj.__class__ == Release:
	projects_to_update.add(obj.project)


	@db.listens_for(db.Session, "after_commit")
	def execute_project_reindex(config, session):
	projects_to_update = session.info.pop("warehouse.search.project_updates", set())
	projects_to_delete = session.info.pop("warehouse.search.project_deletes", set())

	from warehouse.search.tasks import reindex_project, unindex_project

	for project in projects_to_update:
	config.task(reindex_project).delay(project.normalized_name)

	for project in projects_to_delete:
	config.task(unindex_project).delay(project.normalized_name)

	@db.listens_for(db.Session, "after_flush")
	def new_observation_created(_config, session: SA_Session, _flush_context):
	# Go through each new, changed, and deleted object and attempt to store
	# a cache key that we'll want to purge when the session has been committed.
	for obj in session.new:
	if isinstance(obj, Observation):
	# Add to `session.info` so we can access it in the after_commit listener.
	session.info.setdefault("warehouse.observations.new", set()).add(obj)


	@db.listens_for(db.Session, "after_commit")
	def execute_observation_report(config: Configurator, session: SA_Session):
	# Fetch the observations from the session.
	observations = session.info.pop("warehouse.observations.new", set())
	for obj in observations:
	# We pass the ID of the Observation, not the Observation itself,
	# because the Observation object is not currently JSON-serializable.
	config.task(report_observation_to_helpscout).delay(obj.id)


	@tasks.task(bind=True, ignore_result=True, acks_late=True)
	def report_observation_to_helpscout(task, request: Request, model_id: UUID) -> None:

		@tasks.task(ignore_result=True, acks_late=True)
		def update_metadata(request: Request, project_id: UUID):

		@@ -1168,6 +1169,8 @@ def file_upload(request):

		request.db.flush() # flush db now so server default values are populated for celery

		request.task(tuf.update_metadata).delay(release.project.id)

	@pytest.mark.parametrize(
	("states", "exception", "message"),
	[
	(
	[
	{"data": {"state": "PENDING"}},
	{"data": {"state": "STARTED"}},
	{"data": {"state": "RECEIVED"}},
	{"data": {"state": "STARTED"}},
	{"data": {"state": "SUCCESS"}},
	],
	None,
	"",
	),
	(
	[
	{"data": {"state": "PENDING"}},
	{"data": {"state": "STARTED"}},
	{"data": {"state": "RECEIVED"}},
	{"data": {"state": "STARTED"}},
	{"data": {"state": "FAILURE"}},
	],
	services.RSTUFError,
	"RSTUF job failed, please check payload and retry",
	),
	(
	[
	{"data": {"state": "PENDING"}},
	{"data": {"state": "STARTED"}},
	{"data": {"state": "RECEIVED"}},
	{"data": {"state": "STARTED"}},
	{"data": {"state": "ERRORED"}},
	],
	services.RSTUFError,

PEP458: Update TUF repository metadata on project index change #15815

Are you sure you want to change the base?

PEP458: Update TUF repository metadata on project index change #15815

Conversation

lukpueh commented Apr 18, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lukpueh commented May 2, 2024

ewdurbin commented Jun 3, 2024

lukpueh commented Jun 20, 2024

lukpueh commented Jun 21, 2024

lukpueh commented Jun 21, 2024

lukpueh commented Jun 21, 2024

trishankatdatadog commented Jul 2, 2024

lukpueh commented Jul 3, 2024

trishankatdatadog left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

lukpueh commented Jul 30, 2024

trishankatdatadog commented Jul 30, 2024

miketheman left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

kairoaraujo Oct 7, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

This comment was marked as outdated.

miketheman left a comment

Choose a reason for hiding this comment

lukpueh commented Aug 28, 2024

kairoaraujo commented Oct 4, 2024

kairoaraujo commented Oct 9, 2024

Choose a reason for hiding this comment

lukpueh commented Apr 18, 2024 •

edited

Loading

kairoaraujo Oct 7, 2024 •

edited

Loading