GitLab

Quoting RFC: "However, exploiting the case sensitivity of mailbox
local-parts impedes interoperability and is discouraged."

Most of the existing software is threating the local part (i.e. the
thing before @) of emails as case insensitive.

Some people use different capitalization at different times, e.g.
arkadiusz.hiler@example.com with git-send-email and
Arkadiusz.Hiler@example.com with their regular client.

To mitigate this Person's email is now case insensitive for the sake of
searching and normalized to lower case pre_save.

It's not perfect by any means and still can be bypassed by doing
.update() on the QuerySet, but will work in most of the usual cases.

Tests and migration deduplicating existing entries are provided.

Fixes: #32Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

class Series:
    submitter = models.ForeignKey(Person, ...

Would result in Person having a field named series_set QuerySet that
represents the other side of this one-to-many relation.

related_name is used to overwrite that and it looks like it purpose was
misunderstood as it had been set to "submitters" - Person.submitters
doesn't make much sense.

Let's remove related_name an use the default name.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Results send to Patchwork can now be marked as markdown (on a Test model
level). Such results are then rendered into HTML and displayed on the
web views.

Emails that are sent to the mailing list / recipients for such results
are going to be multipart/alternative containing both the text form (raw
Markdown) and the rendered HTML.

Fixes: #6Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Generated mboxes weren't representing newline (aka LF or \n) and
carriage return (aka CR or \r) found in the original patch faithfully.

This is mostly due to how email.Message is implemented - it tries to
normalize newlines in plain text emails. This doesn't work for us as we
want to provide the patches in an unchanged form.

To do that we have to overwrite as_string() and as_bytes() methods on
our Message class that will use modified Generator that doesn't mangle
newlines.

The difference in the checksums is due to slightly different header
wrapping caused by the changes in generator.

Additionally tests are provided to make sure that we ingest emails
correctly (no mangling there) and that the generated mbox is also
a-okay.

Fixes: #33Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

This is done because Python's qp encoding can be a bit problematic with
line endings.

Also, bumping the line length limit to 200 makes the downloaded mboxes
readable with any sensible patches and is well below the RFC 5322 limit
of 998 characters.

The changes in checksums are due to emails not being qp-encoded.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Turns out that clean_header() takes Header, so we were getting:
  AttributeError
  'Header' object has no attribute 'startswith'

Let's use Header in the test and fix the issue properly.

Fix: a7046dca ("Add workaround for MUAs that don't handle UTF-8")
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Noticed it while running tox with the recently added tests.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Looks like it was added untested.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

We already have /patch/msgid/$MSGID to look up a single patch.

This change adds /series/msgid/$MSGID so we can look up a whole series.
This works with patch and cover letter message-ids.

Fixes: #36Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

So people can use it both with and without <>
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Some MUAs seem to send UTF-8 in headers (e.g.: Subject) as is instead of
escaping it using RFC2047.

Those usually get mangled somewhere on the way by the MTA/MDA and we end
up with something escaped with 'unknown-8bit' encoding.

Python raises an exception if we ever try to parse that, so it's better
to at least try to decode it as UTF-8.

Fixes: #43Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

To be used with the recently added event 'until' filtering.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

We have 'since' filtering already and this can be used with some
interesting CI workflows.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

and link to it from the messages about series being incomplete/strange.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

RFC 5322 requires us to not have lines longer than 998 characters and
strongly suggests that we should impose even stricter limitation of 78
characters.

Let's follow that recommendation and use quoted-printable transfer
encoding if any line of payload is longer than 78 characters.
Reported-by: Petri Latvala <petri.latvala@intel.com>
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

We were off by one as the patch's body counts as one comment.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Use the first patch timestamp if available, otherwise use series
completion date.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

When a cover letter is sent as 0/0 it lands in series limbo due to
being in "incomplete" state.

Let's make it visible in the web UI and emit the completion events as
usual, but let's also display big notice what is going on.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Whenever we are duplicating metadata about a revision let's clean the
rerun state.

Without this all revisions to series that had been rerun were stuck in
the "this is a rerun" state, even if updated version of the patches were
sent.

Fixes #31Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Requests generated via `git request-pull` are not meant to be sent in
any standardized way, so users often copy the text to their email client
of choice. This can cause the line endings to be mangled into any
combination of '\r' and '\n'.

Let's be more liberal about what we take with regards to newlines.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Add a button that marks the latest revision of a series as "skip
testing". Only the author of the series and maintainers of the project
can perform this action.

The `skip_testing` state is exposed through the JSON API:
/api/1.0/series/{int}/revisions/{int}/

The attached CI/testing infrastructure is responsible for checking this
flag.

The UI shows which revisions were skipped and who has triggered it.

Closes #27Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Payload is the content of the last text part of the multipart message.

Mailing lists sometimes append their own footers as a new part. This
happens for emails that are already multipart or the ones with more
uncommon encodings.

In those cases we end up with an empty comment, as the footer is then
stripped out via clean_content().

We cannot use the commentbuf as it has anything diff-like stripped out
by parse_patch().

Fixes #39Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

This change will display the # of comments of the patch in patch list
view & series view.
Signed-off-by: Bhanuprakash Modem <bhanuprakash.modem@intel.com>

We are listing msg-ids for patches on their respective pages already.
Let's go a step further and list root msg-id(s) for given series.

root = message that started the series, whether it's the cover letter or
the first patch.

Closes #36Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Old series do not have a completion date for series which breaks posting
results:

  File "patchwork/models.py" in _on_test_result
    1083.                   'revision_completed': revision.completed.isoformat()}

  Exception Type: AttributeError at /api/1.0/series/71839/revisions/1/test-results/
  Exception Value: 'NoneType' object has no attribute 'isoformat'

Let's just admit that it's unknown.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Test results are listed on a series view. Some of them are expandable if
you click anywhere on their grey background.

This is not obvious and it's not indicated in any way.

This patch adds a triangle pointing down in the left corner to imply
this expand/collapse behavior.

The triangle starts pointing up when the section is expanded to indicate
that it is collapsible now.

Closes #35Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

The event is logged when patchwork receives new test result for series
revision or for a patch.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Previously there was no way of telling when the series is complete other
than digging through the completion events, which is suboptimal.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

This solves the following error:

File "patchwork/views/patch.py" in patch
  82.                              % (patch.name, bundle.name, ex.message)]

Exception Type: AttributeError
Exception Value: 'Exception' object has no attribute 'message'
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Fixes: c626d5ea ("Add robots.txt")
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Disallow all crawling.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

When running "git pw apply --revision N SERIES_ID" it always says
that it applies the newest revision even though it does the correct
thing - it applies the requested one.

This patch fixes the logic to print the correct revision.

This fixes #22Signed-off-by: Bhanuprakash Modem <bhanuprakash.modem@intel.com>
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Let's address a common gotcha - clicking on a project name in the navbar
used to navigate us to the main page (i.e. `/`), since this patch it
will get us to the series list of the current project instead.

This fixes #30.
Signed-off-by: Manish Kumar <manish3.kumar@intel.com>
Reviewed-by: Martin Peres <martin.peres@linux.intel.com>
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

The all-new path() does not take regexps.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Wrong view, wrong way of passing view arguments.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>