The new API requires at least one event and will automatically save a
snapshot of the document and related state. Document.save() will now
throw an exception if called directly, as the new API is intended to
ensure that documents are saved with both an appropriate snapsnot and
relevant history log, both of which are easily defeated by just
calling .save() directly.
To simplify things, the snapshot is generated after the changes to a
document have been made (in anticipation of coming changes), instead
of before as was usual.
While revising the existing code to work with this API, a couple of
missing events was discovered:
- In draft expiry, a "Document has expired" event was only generated
in case an IESG process had started on the document - now it's
always generated, as the document changes its state in any case
- Synchronization updates like title and abstract amendmends from the
RFC Editor were silently (except for RFC publication) applied and
not accompanied by a descriptive event - they now are
- do_replace in the Secretariat tools now adds an event
- Proceedings post_process in the Secretariat tools now adds an event
- do_withdraw in the Secretariat tools now adds an event
A migration is needed for snapshotting all documents, takes a while to
run. It turns out that a single document had a bad foreign key so the
migration fixes that too.
- Legacy-Id: 10101
Used multipart mime when including the original message in an error message to the secretariat.
Modified the way IANA review email is parsed.
Added non-ascii UTF-8 tests to the IANA review email parser and to the SMTP Exception handling code.
Commit ready for merge.
- Legacy-Id: 7673
(-r5194:5465 from branch/iola/shimfree). Copying relevant
commit messages here:
- Removed .related many to many relationship, it's not really useful
since we always have to restrict on the relationship type anyway,
instead add two helpers for doing the necessary queries (in both
directions)
- Added migration for transforming the .desc on the new_revision events
into something more akin to what is actually shown in the history page
- Added migration for blanking IESG notes that just consist of "RFC
XXXX", these have been superfluous for some time
- Grant stream chairs access to changing the stream on a draft
- Hacked the format_history_text filter to be less weird, using the same
formatting for snippets and full text, also link up legacy ballot set
events
- Moved the decoraters + utilities to new ietfauth/utils.py file
- Added simple helper to Email to identify invalid email addresses (from
legacy author entries)
- Used new new_revision .desc format for when drafts are submitted
- Improved the looks of the button class by adding extra contrast and a
linear gradient. Currently the gradient is only visible in fairly
recent browsers.
- Rewrote draft and RFC tabs in terms of the new schema, porting
write-up and history tabs as well
- Fixed two bugs in RFC Editor syncing: make sure documents we don't know
beforehand get a "draft" type and make sure individually submitted
drafts get the type="individ" group instead of NULL
- Made the CSS-styled button feel a bit nicer to use by flattening the
active state, also introduce some temporary styles until browsers
catch up with the standard syntax
- Added migrations for fixing 1) a dummy RFC entry, 2) three stand-alone
RFCs that didn't get their doc.type set, 3) a big bunch of historic
stand-alone RFCs that have doc.group=None - set these to the
individual submission "none" group for the time being so the view code
doesn't have to deal with a special case.
In some cases this is wrong since there actually was a WG associated
but unfortunately fixing them properly requires detective work
(probably parsing the RFCs) and in at least some cases recreating
historic WGs. In case someone ends up doing this, the documents to
check can still be found with
Document.objects.filter(name__startswith="rfc", group__type="individ")
since there are almost no new RFCs that didn't went through the I-D
process.
- Merged the I-D and RFC views by showing I-D information on RFCs too.
I-Ds that have been published as RFCs redirect to the RFC URL. Also
support alias URLs so e.g. /doc/bcpXXXX redirects to /doc/rfcXXXX.
- Fixed revision augmentation so events after RFC publication gets a "RFC"
designation
- Fixed a bug with tabs not using provided name but rather doc.name
- Displaying draft-iesg state rather than doc.friendly_state as IESG state,
also show a notice that the IESG state refers to post-RFC processing
if it does, like the old separate RFC page did
- Fixed the RFC number doc.note migration to catch combined "RFC XXX; BCP
XXX" notes too, use the opportunity to remove inserted HTML tags from
notes and rely on linebreaksbr filter instead (the other thing was a
left-over from the Perl days), update the various uses of the note to
reflect that
- Refactored slightly to make views_doc.py independent of other idrfc code
- Moveed idrfc/views_doc.py to doc/ with associated templates, replace the
somewhat fragile simple URL tests for views_doc.py with ordinary unit
tests. The new tests are still fairly basic but at least test more
than the URL tests did.
- Made sure RFC's (and BCP/STD/FYI) are stored as RFC123 instead of
RFC0123 in the alias table with a new migration and a change to the
RFC Editor sync, this in turn makes /doc/std1/ do the right thing
- Now /doc/std1/ works, we can actually do a local link in
urlize_ietf_docs rather than linking to the tools.ietf.org server
- Fixed history text formatter: sanitize HTML before adding linebreaks and
non-breaking spaces, this cuts the time to render a history page with
long comments in half
- Added a test crawler that walks through the crawlable part of the site,
reporting errors and slow pages
- Got rid of initial "No record" positions when showing old positions,
it's just noise
- Added a .select_related() to the document main tab to reduce the number
of DB queries, unfortunately it seems it doesn't really help with
Django 1.2.x due to a bug (Document inherits from DocumentInfo which
makes things a bit more complicated)
- Introduced a simple cache in doc.get_state so repeated reads don't
cause a DB query
- Cleaned up the search code in preparation for removal of the shim-layer;
use a static button and don't send extraneous GET parameters
- Removed dead code in several places
- Legacy-Id: 5830
from IANA as it turns out that the Datatracker is the authoritative
source on these states. Also improve logging so that the raw JSON from
IANA is dumped, the parsed JSON on imported changes is dumped and we
write to syslog before starting a sync script in the notification view.
- Legacy-Id: 5800
discrepancies report for seeing differences between the
Datatracker/RFC Editor/IANA, add tests of IANA/RFC Editor integration,
add script for weekly discrepancies emails
- Legacy-Id: 4851
