The correct behavior was depending on the order in which swh.web django
apps are registered. This should allow to have less entries in the
SWH_BASE_DJANGO_APPS so we can have an instance only serving /api for
example (which is currently not possible).

These default config values are mostly irrelevant for any kind of
deployment. Removing the default values make it clear and mandatory
these need to be explicitly configured.

This also prevent the swh.web.utils.archive module from
executing (backend instantiation) code at module loading time.

Introduces a new patch_backend fixture to help mocking a specific method
of a known configured backend (storage, vault, search etc.) to try to
provide a standard and unified process to mock these in all tests.

remove SWH_WEB_XXX_NAMES constants as well as 'xxx_server_names' config
entries. These are not used any more.

Add a new 'corner_ribbon_text' config entry to specify the text displayed
in said corner ribbon. It should allow to remove some very swh
deployment specific hardcoded constants (especially urls).

It makes tests a bit easier to read and maintain.

Replace usage of the hardcoded and swh specific
SWH_WEB_INTERNAL_SERVER_NAMES by a more generic UNAUTHENTICATED_HOSTS
config option to bypass authentication for the graph public APIs.

This uses the same validation logic as django's ALLOWED_HOSTS one.

Antoine Lambert authored 3 months ago and Antoine Lambert committed 3 months ago

Network issues might happen that can slow down the loading
of README byes so display a spinner by default in the card
holding README content.

Fixes #4827.

Some subversion repositories use these credentials for anonymous access
so allow to submit origin URLs with such credentials in Save Code Now
form and Web API endpoint.

The filesystem of a revision is browsable from the associated view
but the display of the languages selection dropdown was missing when
browsing a content.

Ensure languages selection dropdown is displayed for content whose
mime type starts with application/ if the content encoding is not
binary, for instance application/json.

The forcing of visit type introduced in 937a9609 should only be
done when no snapshot related info have been passed as parameters
to the get_snapshot_context function.

Scheduler temporary backend has been removed in favor of a more
efficient memory backend.

Antoine Lambert authored 1 year ago and Antoine Lambert committed 4 months ago

When no visit type for an origin is explicitly provided as query parameter,
ensure a full VCS visit type (git, svn or hg) is selected by default when
an origin has multiple visit types.

Related to #4786.

Previously, the length query parameter value sent to datatables list
views implemented backend side was not restrained to a maximum value,
making a potential DoS vector attack.

Set that maximum value to 1000 which is the max page length offered
by datatables Web UIs in swh-web.

This maximum value can also be modified through configuration.

Fixes #4820.

When the path query parameter targets a content in the root directory
and has a leading slash, a 404 error was previously raised so ensure
to handle that special case.

When browsing an archived object without any context (a plain revision for
instance), the javascript variable swhid_with_context_url is null so the
code computing parameter for the Web API citation endpoint was raising an
exception.

Spotted thanks to the cypress tests execution on Jenkins.

Django unescapes percent encoded query parameters when parsing them so we
need to ensure the % character a SWHID can contain is escaped to avoid
parsing an invalid SWHID.

The minify-html tool use to minify HTML through a django middleware
strips the type="submit" attribute from button elements so use ids
instead to get such elements in cypress tests.

Use HTML minification as in production to catch possible issues.

Our custom HtmlMinifyMiddleware was using htmlmin package under the hood
but it is no longer maintained and does not support Python 3.13.

Prefer to use the django-minify-html package instead which is maintained
and offers better performances as it is using an HTML minifier written in
rust in its implementation.

Canonicalise only forge URLs identified by pabs and not the whole
set in database.

Really fixes swh/infra/sysadm-environment#5504.

Fixes swh/infra/sysadm-environment#5504.

Some tests were using canonical Bitbucket/Gitea/GitLab domains that
should be blocked from being submitted to Add Forge Now.

Other tests were using domains that have been parked.

Otherwise it could match URLs containing GitHub in the path/search/anchor,
or with domains similar to GitHub ones but without the periods.

Antoine Lambert authored 4 months ago and Paul Wise committed 4 months ago

Check AFN creation form is rejecting new types of invalid forge URLs
according to newly introduced forge URL validation rules.