It's possible for some load failure states to cause this teardown at a point
without a surface. I observed it in MobileFrontend, when restbase can't be
contacted.
Change-Id: I13489640759c73dad5730d0bbadbfb2a38cb8e53
This lets copy-paste between documents retain the numbered status rather than
falling back on pasting "<a>[3]</a>".
Update the part of LinkCache which selects on mw:ExtLink, so it will handle
possible multiple values in the link rel.
Bug: T188429
Change-Id: Ia5e4c9fa45e94da9cbfcd2a42d017d0fda1c511f
Always store immediately if fromEditedState is true. Also
now that we only store if there is state to recover, remove
the check for transactions before deciding to show the notification.
Change-Id: I5357a9098b91e303f5c71881ea03a080d2969fff
Specifically, set arabic(extended) and hebrew to 'rtl'.
Logically depends on I14abd3e0c0d23f79aa01d96c216eea913024b4c8
to set the dir attribute in UI.
Bug: T56310
Change-Id: I1c7e28d3d2f20ca84115be6d49650cd9a81d78dd
1. It wastes valuable time during setup.
2. If a user reloads the page without making changes we
should give them the latest html from the server to
minimise the chance of an edit conflict.
Change-Id: I9a1f8cfd65ef2552fe2c3d6d2bbf975851b52003
Ensure we start with the same HTML (i.e. if an edit has
been made since the crash-recovery):
* Whenever an article target is activated, stash the initial
document html, other parsoid response data, and the request
parameters (pageName, mode, section) in session storage.
* Whenever an article is fetched through the target loader,
recover from session storage if the request parameters match.
Store transactions:
* On document transaction (debounced) append the latest
changes to session storage.
* If a document state is recovered from session storage,
attempt to re-apply the stored transactions.
Clear transactions:
* Whenever the target is torn down (i.e. save, deliberately
closing the editor to go back to read mode)
Other:
* If writing to session storage fails once, disable future
attempts for that session (assume storage quota exceeded)
* Disable tempWikitextEditor when recovering. We don't have
the transaction code loaded yet to perform the recovery.
Bug: T57370
Depends-On: I3832243fc347a99641fcb7e39a887a153c9a3b22
Depends-On: I448fb566fe9f7f5b5a76e88b70ca000e3d35b415
Change-Id: Id9d877f903cf4796a52f90991c030417a9f8786f
Allows users to know when the widget has been constructed,
and access it (e.g. to set an initial selection)
Bug: T185279
Change-Id: I3678996bcf644cc889dd168ac3ce48b5c3633ec1
This line isn't solely for supporting FF52, that is the order
in which it is called (move after attach, not before),
but that matches all our other widgets, so not sure
it needs commenting.
Change-Id: I6f3cc5687f1e4b995dff700d0765d14de1927d51
mw.storage catches errors, so we won't crash horribly when the user has
localStorage disabled / full.
Bug: T181822
Change-Id: I212994eb535b9a8fb5f6c09deaa10b16c3d7f10e
Accesskeys in this situation are unreliable cross-browser. Firefox won't
trigger the accesskey inside the dialog. We can manually simulate it, via the
existing trigger.
Bug: T121183
Change-Id: Ib919d8b9fcd9324a517037bcc6ef93a26d1488b9
We emit <span id="..." typeof="mw:FallbackId"></span> inside headings
when necessary to provide backward compatibility with pre-HTML5
section ids. These are mostly harmless inside VE --- they become hidden
alienated nodes, which are then safely ignored by Parsoid during html2wt
even if they manage to migrate outside of a heading --- but it's a
little safer to strip them.
Change-Id: I07d0fcf54427ab02d4930cff183d3e5579a99306
Tell browsers we explicitly want spellchecking on the section title. If
unspecified this is browser-dependent, so it was inconsistent.
This won't actually have an effect until the next OOjs-UI release (v0.24.3),
as this was a new feature added to TextInputWidget.
Bug: T179654
Depends-On: Ifdd44ead7d782288aaef0979322f0943f7304835
Change-Id: I588874efcff91c5be03da2fc03c268f883b718e2
We need this hack when visually editing articles, but not in e.g. the
Flow integration. Move it down so that mw.Target has a clean
parseDocument method that can be used by Flow without undesirable
newlines being introduced.
Change-Id: I52bba88b3d33c3d6d16c4cf832f23f970a243abf
Idea being: preload use case is often "load content, maybe with parameters,
immediately save it", so we can treat it as being already-modified for saving
purposes.
Bug: T179063
Change-Id: I35ff04bc9cb5172edf8174832624145973aa1fdb
We were reimplementing (incompletely) the setup from the .init version. Just
call the original, modified slightly so it doesn't over-setup on repeated
calls.
Bug: T151021
Change-Id: I65bd7c5ecf75c478d6babeb13e7fb2a76a9842c8
The whitelist isn't checked for blacklisted browsers,
so if all non-blacklisted browsers are whitelisted,
set the whitelist definition to all (null).
Change-Id: I8ee3569d9d5be2eda9153f1b087c1be385f93d9c
ve.init.Target sets up TriggerListeners based on getSurfaceConfig if there's
no Surface instantiated. As such, loading directly in source mode would have
those listeners present and assuming the document is in visual mode. So, if
there's not a Surface already, we'll assume we're going to get one created
with the defaultMode, and set things up expecting that.
Bug: T153959
Change-Id: Ia0afe32e6b63ea0d3418137ca5c8bc32c009ecfa
* Rename 'deactivate' to 'tryDeactivate' as it may prompts
the user to deactivate.
* Merge 'cancel' and 'teardownSurface' in to 'teardown',
extending the parent method.
* Rename elementsThatHadOurAccessKey to $saveAccessKeyElements
and move teardown to parent class where it is setup.
* Move toolbarSaveButton teardown to parent class where it is setup.
* Cleanup changeDocumentTitle
Depends-On: I9d97614695272dca6936ef6f3461178fcf0368a8
Change-Id: Ie998a04c21f6615b4415edf471310db5edca3b5a
This code is meant to fix up the tabs if wgVisualEditorTabPosition
was changed and we're seeing cached HTML with the tabs in wrong order.
But it seems it has never worked. `$caEdit[ 0 ].nextSibling` is a text
node and obviously not equal to `$caVeEdit[ 0 ]`, and vice versa for
the other case. The logic is actually correct, so let's fix it to skip
over text nodes.
Also, remove a stray 'eslint-enable' comment left over in
a0f934ed26.
Bug: T50017
Change-Id: I23663a9bfcfdbf52918452c878a128e6960b1191
Vector no longer does the silly thing where the tabs are reversed in
the HTML in RTL languages, so we must remove our hack to do the same,
since otherwise it reintroduces the issue it was intended to prevent.
This essentially reverts 2efd4f0061.
Bug: T50017
Change-Id: Ibeaa55bc34899bffab849ea8ad8b127fb5184d43
The default value of target#pageName is wgRelevantPageName
but other targets my override this, or change it dynamically
(e.g. ContentTranslation).
Also remove duplicate setter of pageName in mw.ArticleTarget,
already set in mw.Target.
Change-Id: Iebd1def1d4142978a673afec584a0b663644d176
* Tiny fix to 33dc60838c for PHP variable name
* Bigger fix to properly pass preloadparams in and split them up
Change-Id: I844db115f2563cb9ee1629c30d5f49d1ce58f5bd
This allows the use of the preload and preloadparams query parameters. They
should behave as they do in the old editor, loading substituted content in
visual and source modes.
Bug: T51622
Change-Id: I522fb5b480d17912f6d6116be6aa043ead855b52
Other page state is read directly, rather than passed in.
If we come up with a use case for passing this in at a later
date we can reconsider the best way to do this.
Change-Id: Iee7e608c969776b89dbb38b3b56a5177e48344af
Instead, use getCheckboxesDefinition(), and build our save form
checkboxes from that rather than extracting them from the HTML.
The ability to have non-checkbox fields there is removed, as that was
never intentional and is now impossible.
To avoid transient problems during deployment (old JS code cached in
the user's browser receiving the new format of API responses), the old
property is kept in the API response. We should remove it next week.
Bug: T174613
Bug: T174686
Change-Id: I5bfca5e116fe790302c3b6ac1357e80237fb1ed2
ve.init.mw.Target.static.parseDocument requires that its parameter be
a complete HTML document with a single root tag, otherwise some of its
workarounds break horribly in IE 11 (and possibly Firefox).
Bug: T171600
Change-Id: Iaacf24f7bef40a24230fcfa08e15bab31b1587c0
Implement special node types for language variant markup, so that they
display appropriately based on the currently-selected variant.
(Parsoid uses empty elements to represent this markup, so without this
patch anything in -{ ... }- is alienated and disappears.)
A follow-up patch will implement context items and inspectors to
allow editing these nodes. This patch is basic "read-only" support.
Depends on I4fcdebc2290ec35ba188f4c2e69d578791fbcd67 in Parsoid to
generate the appropriate markup, but this patch is safe to merge
independently.
Bug: T49411
Change-Id: Ie11e9301d2513bfe4a36036481cee9a047f46d37
This stops the welcome dialog from being displayed repeatedly if the user
edits the article multiple times in the same pageload.
Bug: T170983
Change-Id: I8cc5990e9acb1ac09a2d6dd95eef8042c70088c7
Selflinks with fragments aren't actually selflinks, and so shouldn't be styled
as such.
Bug: T170943
Change-Id: Icb67c81327769d09af6bcfe593843d9dd2bcc33c
This also removes an undeclared dependency on mediawiki.Uri.
Bug: T169574
Depends-On: Ic4abaae6e9c74509f7a114a697cdbb78dd1b89b6
Change-Id: I76803da5f1a52171a1af8de1f3e31ea127a2da78
"Welcome to wikipedia, anyone can edit, we welcome all improvements. Start
editing!" is a bit out of place when what you'll see after clicking "start
editing" is "you can't edit this page".
Bug: T138715
Change-Id: I9f655a5f12d4e45644bd01631c2d3131375d8e8f
A missing image should return some data to say the
image is missing, not just reject the promise.
Bug: T169337
Change-Id: Ib41a64a783c1baca88f428417c98e7fb913d14a1
Only shown if VisualEditorEnableDiffPage is set, or query
string param 'visualdiff' is present.
Currently:
* All VE javascript is loaded (could be cut down to just DM code)
* The entire Parsoid HTML of both revisions being compared is loaded
* Both Parsoid HTML docs are parsed into VE DM trees and diffed
Bug: T167508
Change-Id: I151fc9bab3d3032f50c8d11be6b54e45a06fcc34
No longer needed after I2ede6c88be569224c0b2a4e388a133ad98b86a7b
in VisualEditor core.
* 'transform: none' from 049a32c60a
(that change does not describe what the issue was, but clearly
it wasn't a complete fix and I see no problems without it).
* 'z-index: 4' from 6fff13b898 (I can't
reproduce the issue there after removing it).
Bug: T167616
Change-Id: I7e0d603af2668269e5587eaa5e4761166a1b4097
Add 'require' to eslint global whitelist for this extension,
since ResourceLoader provides this in all contexts now.
Bug: T166254
Depends-On: I0e6a95bd4fbaba06eae5137ac7af84f62bebf6f0
Change-Id: I6b2fe57317dcff18ec6d9ea0cd9e4eb9c894c169
Any MW target may want to get a wikitext fragment, not just
ArticleTargets. Requires us to move refreshEditToken up as well.
Change-Id: I27d30e20fefd5ba266d3d8e5f99383e9b4d4993a
Follow-up to fefb76eebc. Prior to that
change, the condition for this looked like this:
// … if on a ?veaction=edit/editsource page
(
isViewPage &&
uri.query.veaction in editModes &&
(
uri.query.veaction === 'editsource' ||
init.isVisualAvailable
)
)
In the refactor, the `uri.query.veaction === 'editsource'` check was
lost.
Since that code is pretty messy (probably predating the source editor
and hastily adjusted), instead change the check for `isVisualAvailable`
to just `isAvailable`. If the requested mode turns out to not be
available later, the editor will not load.
Bug: T165146
Change-Id: Idfaf9115dd20cec8f8e044a704b93b07984cdcee
Ideally we could use the preferred editor, but this breaks
tab switching to the old editor.
Bug: T165238
Change-Id: I5f5ee5566cdd2080ba7c89d43cf127b457537768
We added support loading NWE from action=edit in I35208cce069
but missed this check in the front-end.
Bug: T165238
Change-Id: I2732eaa81a3f968b34c4e878b2ad36de981dd567
Using the forthcoming wgRelevantPageIsProbablyEditable.
Bug: T165010
Depends-On: I6c6ca1cfd93e7be917952980f1e1d57aec3a1292
Change-Id: I6c6ca1cfd93e7be917952980f1e1d57aec3a1229
* Fix focus and resize/scroll into view after the
captcha has loaded (so after the image has loaded
for FancyCaptch).
* Add margin between input and captcha
* Enable enter-to-submit on captch input field.
* Unify code for simple/math/questy and correctly
render math as HTML (by looking at the mime type)
Change-Id: I10433cefbfea8569674c120dde5b489570e20966
* Do not try to load the editor on protected pages, or if the user
doesn't have permission to edit pages.
* Move the check for required DOM elements after the check for
pages without the editor (T162411) and after DOM ready (T163307).
Bug: T162411
Bug: T163307
Change-Id: I8149694ba8155682701f2cda6ca212d60f446caa
Expanding from a single if-expression lets us make the
code much more readable, and reveals some minor bugs.
Change-Id: I49e57bfc093e019c837a73eab5c25fdbd14de0af
Follows-up 914eb1cde3; reverts the logic to have the label change, but
sticks with new "-current" i18n messages, so the "X editing" instead of
"Switch to X editing"; the old i18n messages are now dropped entirely.
Bug: T162864
Change-Id: Icdad6c5bb06ff86e9b8976840bb1a94b939609a8
Sites like to customize 'readonlywarning' with things that we can't parse.
Instead, use 'apierror-readonly'. The old message is still shown in edit
notices.
Bug: T163455
Change-Id: I592accde17b256ecee27820df12d9d312499f8f7
If we retry and get another 'badtoken' error, something is very wrong,
and trying again and again is probably pointless. This behavior also
matches how mw.Api#postWithToken handles bad tokens.
Bug: T163493
Change-Id: I488fd7e74e37922044491695d6c9025a45531281
Previously we completely disabled the loading of notices
when switching, instead of just not showing them (so they
don't clash with the "switched" popup).
Bug: T162812
Change-Id: I3f8e787630e196cee1dbb1aa449b3558b74fcd04
New changes:
57f8064b6 tests: Enforce dm unit test coverage at 20% per file and 80% globally
6f757c670 LinkAnnotation: add a getFragment method for consistency with MWInternalLinkAnnotation
c40a3f152 Drop the classList polyfill for IE9
d1af6c6bb Update OOjs UI to v0.21.1
bfb35aaef Localisation updates from https://translatewiki.net.
Bug: T162277
Change-Id: I5146bc421bb88e56bafadccefc7bbdee106504ea
This is run after the targetPromise completes (via the platform-agnostic
setDefaultMode proxy), which is the correct time to do so. We should not
do this twice, and we definitely shouldn't do this before the target has
successfully loaded, lest a user ends up with us setting their option to
"prefer" an editor that they actually can't load. Whoops.
Bug: T156316
Change-Id: Icf4b5ddd9c8265ade55f43328f807344b41db350
* Pass a wikitext promise, instead of wikitext
* Handle writing of content to the dialog from within
the dialog.
* Handle diff errors within the dialog.
* Never disable the 'Return to save form' (approve) button
* Remove redundant messages
Change-Id: Ibd76e8951998f751abfb4f407682202c2f73ac7e
New changes:
dc0cad28c tests: Factor out ve.dm.TestRebaseClient/Server into their own files
98061108b Convert RebaseServer tests to data provider
3e746a51b Update OOjs to v2.0.0
16231a812 [BREAKING CHANGE] Drop support for Internet Explorer 9
64daeb4b5 Localisation updates from https://translatewiki.net.
3310e876c Update DOMPurify to 0.8.5
579033a83 Follow-up 16231a81: Drop IE9 support in createDocumentFromHtmlUsingIframe
Bug: T162277
Change-Id: If17c6a3fac0920db29f80069f344ffd6be46c3c9
This will be required later when we have to fetch the original
document from the server before running a visual diff.
Change-Id: I351688f4d1fec892bc297f40c209939cbb2ccf95
Since the PopupWidget rewrite, margins for aligning popup anchors
should be set on the anchor itself, not on the popup.
Change-Id: If8237ae5edbc2e8c5cb99147f392aae52e642765
The styles are not specific to the Vector and MonoBook skins, but rather
to the MediaWiki and Apex themes of OOjs UI. Rename them to reflect that.
Change-Id: I757e8d96625feb8290cbf7837b5d801961cda344
Provide a utility funcition in ve.init.mw.LinkCache to do this.
Also use this in ve.ce.MWTransclusionNode/ve.ui.MWPreviewElement, and
introduce to ve.ce.MWExtenionNode
Bug: T73900
Bug: T153535
Change-Id: Ieb9a0274b8c5ae1932c431546f09d18000fa6dd9
This is a little inelegant, but it works in both CE and transclusion
content.
Bug: T52497
Depends-On: If058843924c3b30c116df2520aef93a004d98a5d
Change-Id: I88d45aad2aaa45e71b433350986b19764603a1f2
The bare minimum of elements that must be present on the page for
VisualEditor to start and mostly work is:
* '#content' to attach our interface (toolbar etc.)
* '#mw-content-text' to replace with the editing surface
* '#ca-edit' to start up the editor
When those elements are missing (and we expect them to be present,
e.g. we're not on a special page), we print a console warning.
For example, VisualEditor now launches on Timeless and CologneBlue.
It does not launch and prints a warning on Modern and Erudite.
(It is actually possible to edit and save pages on CologneBlue,
although various pieces of the interface look funny.)
This does not change which skins we claim to support: these are still
currently only Vector, MonoBook and Apex. But it allows third-party
skin developers to more easily test the editor and fiddle with their
CSS to make it look right.
Bug: T161373
Change-Id: I97c786f3c8c795c238b10bef332a129ea26f86c2
The element has both the class and the id set to 'catlinks'. This is
just for consistency, I have no preference on which to use.
Change-Id: Ic96b0f21fe790d0aa03bfd6366ba246e8493a038
Redirect pages have two extra things not present on normal pages:
* Redirect subtitle: the "Redirect page" shown under the page heading.
* Redirect page content header: the "↳ Foo" at the beginning of content.
Our handling was pretty messy and had some bugs (T161614).
Notable behavior changes:
* Update 'wgIsRedirect' in mw.config after saving the page.
* Use #mw-content-text rather than .mw-jump for inserting the fake .redirectMsg.
.mw-jump is not guaranteed to exist on the page (it's a skin feature).
* Never replace the real .redirectMsg existing on the page, except by the
new HTML after a page is saved. Our fake is separate now.
Bug: T161614
Change-Id: I96a5e45a71bf10bf6a2b501dc0cf81e6c37060ec
To reproduce: start editing a page, turn it into a redirect,
cancel editing and go back to view mode. The "Redirect page"
subtitle under the page title should disappear.
Note that #redirectsub is correctly spelled all lowercase,
unlike #contentSub.
This reverts commit 0e1bc7309b
and fixes the code instead.
Change-Id: Ibacd73122dfec63268a77794bc77c4b88876d3ee
As it happens, #redirectsub is correctly spelled all lowercase,
unlike #contentSub. This line does nothing. We remove #redirectsub
correctly elsewhere; I don't think this was ever meant to be here.
Follows-up 96421b283c.
Change-Id: I3e4c6eb2ff94f363b488477b3ddd248e571e723a
