GitLab

If navigating to a patch from a series/revision view a ?series=x&rev=y
are appended to the URL.

This is later used on the patch view to determine whether browsing this
patch from the series view. If that's the case new navigation options to
go to next and previous patches in series are available.

Closes #13Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Complementary to /mbox/ and /mbox_with_cover/.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Add id to comment divs and turn the dates into a hrefs those anchors.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

When a comment containing a diff is sent to the mailing list is has the
diff extracted out and we save just the comment body above the patch.

Let's fix that by using payload instead of commentbuf. We still strip
signatures and mailing list footers there.

NOTE: we still use commentbuf for splitting an actual patch into the
diff and the patch's comment.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Currently each time there is a comment with a duplicate msg-id, ADMINS
get an email with an exception.

Let's be more graceful and just debug log it.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

FDO started respecting DMARC, which means it cannot freely resend emails
without rewriting 'From:' header in some cases.

So this:
  From: Author <original@mail>
Changes to:
  From: Author via listname <listid@lists.freedesktop.org>
  Reply-To: Author <original@mail>

Let's recognize that and unmangle it for authorship and mbox purposes.
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 clarify descriptions of the other options.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Let's assume mail structure with replies like this:

[1/2] Patch 1
  |- [2/2] Patch 2
  |- [2/2] Another Patch 2

When the "Another Patch 2" comes in, it's going to get inserted in the
same revision at the same place as the original "Patch 2", which results
in integrity error.

This is the single most common reason for administration notifications.

There is not really much we can do, as the reasons for such emails are
so ambiguous that the best we can do is to discard them, just like we
are doing now. So let's make it silent.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Couple of small issues like old syntax for print with stderr and
comparing numbers using "is" instead of "==".
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

So the patchwork stands out a little bit more among all the opened tabs.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Due to complex handling of incoming series all the new cover letters
were not used.

parsemail creates new, transient revision that is later discarded.
Everything that should get persistent in the proper revision has to be
copied over explicitly in on_revision_complete signal handler.

So let's do that for all the cover_letter fields, but only if they are
present.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

If cover letter looked like a patch (i.e. had the diff markers) it was
parsed as such, resulting in the diff being extracted out from the
contents and stored as oddly numbered patch 0.

Let's store the raw payload and do not create patches out of cover
letters.

This is now tested by adding some diff excerpt to one of the test
fixtures and adjusting the asserts on cover_letter and raw_cover_letter
accordingly.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

To not waste time on reinstalling all the deps each time
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Parallelize runs, use in-memory sqlite3 and turn off unnecessary
debugging/logging.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

We were testing for that, but yet we haven't expressed it in the code,
relying on the default behavior of the database. This broke sqlite
compatibility.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Project and Person instances were stored globally, keeping their PK(s)
even when db cleaning removed them, causing some saves() to overwrite
them when used with sqlite3.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

They don't affect passing on any of the 3 major databases in multiple
runs, so it seems that the issues that the sleeps waere trying to
mitigate (ordering) was fixed elsewhere.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Be more explicit about how we sort when resolving project and don't
discriminate different ways of sorting empty string vs blank chars.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

 * Use django-rest-framework views for nice browser previews
 * Add pretended pagination for consistency with other list endpoints
 * Extra: remove accidental print
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

The only difference in comparison to patch is a lack of patch_id in the
results.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Cover letter is also a part of revision "metadata". Coping it over makes
the overall experience better - all the meta tags included in the cover
are preserved and user can see it on revisions caused by a rerun or a
single patch bump.

The difference between duplicate_meta and duplicate is that duplicate
also copies over patches, this may need a better naming.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

`/api/1.0/series/$series_id/revisions/$version/mbox_with_cover/` has been
added.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

We are now storing unprocessed cover letter and it's headers and
exposing it through the API.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

With this change hash part of the URL is updated when clicking on a
revision to show.

Also providing hash in the URL will navigate straight to that revision
(e.g. /series/1234/#rev3).
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Simply reload page on rerun, so the newly created revision is visible
and all the dates are updated.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

the endpoint is: /api/1.0/msgids/$MSG_ID/
msg-id works both with and without <>

output:
[{'patch_id': 112223,
  'project_id': 9,
  'series_id': 12938,
  'revisions': [1,2]},
 {'patch_id': 112224,
  'project_id': 18,
  'series_id': 12937,
  'revisions': [1]}]

The implementation is not using django-rest-framework because the
generated routes disallow dots in the identifiers and requires adding
much more boilerplate to have it working.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Mail object append headers, so to overwrite them correctly we have to
delete the original one first.

Also fix a chicken and egg problem spotted by this fix.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

The regexp breaks with patch series having more that 9 patches, because
'*' was a greedy qualifier, eating up every but the last digit of the
patch number. I have forgot to commit the extra '?' which makes it
non-greedy, making the whole thing behave as expected.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

Not only it was unused but also used for something else than the name
has indicated.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

For simple [PATCH xx/yy] we store just [xx/yy] but for more complex
[PATCH extra xx/yy] it becomes [extra, xx/yy].

Let's fix the regexp used for extracting patch numbers, so that extra
tags do not make revision strange.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

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

Include information that series is not complete (i.e. we have received
less patches than expected, basing on the [xx/yy] tag) and that series
looks "strange" (i.e. ordering may is off or we have got too many
patches).

Changes in logic:
If we get more patches than expected, *do not* reset status to
incomplete series, we've got the event for queued already anyway.
Just consider it "strange".

Also, let's include information which revision we are rerunning on the
rerun button.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

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

Up until this patch, rerunning was just creating a 'newrevision' event
for the existing revision, effectively queuing it again. Any incoming
results would overwrite the existing ones.

Now, whenever rerun is hit, the last revision will be duplicated (no
results overwriting!), bumping "Last Update" time for the series. We
also mark such series as a rerun so we will be able to display it for
the users / expose it in the API.
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

A lot of minor changes.

W504 is getting ignored, as we break the line after the binary operator
throughout the code base (alternative would be W503 - break before the
operator).
Signed-off-by: Arkadiusz Hiler <arkadiusz.hiler@intel.com>

patchwork returns a following back-trace without this patch:

Internal Server Error: /patch/858/
Traceback (most recent call last):
  File "/usr/lib/python2.7/site-packages/django/core/handlers/base.py", line 132, in get_response
    response = wrapped_callback(request, *callback_args, **callback_kwargs)
  File "/srv/patchwork/lib/apache2/../../patchwork/views/patch.py", line 34, in patch
    editable = Can(request.user).edit(patch)
  File "/srv/patchwork/lib/apache2/../../patchwork/permissions.py", line 38, in edit
    can.edit(patch.project)))
  File "/srv/patchwork/lib/apache2/../../patchwork/permissions.py", line 32, in edit
    project in self.user.profile.maintainer_projects.all())
  File "/usr/lib/python2.7/site-packages/django/utils/functional.py", line 226, in inner
    return func(self._wrapped, *args)
AttributeError: 'AnonymousUser' object has no attribute 'profile'

Fixes: e876b068 ("permissions: Introduce central point for defining permissions")

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

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