The documentation on the matrix settings is particularly poor.

Failing unit test submitted to Charles Jolley

In Tasks sample app, if you update a Project name, all Tasks are fetched - should only update Project record instead

Back in Quilmes I did a context.setClass('custom-view', YES) in my renderer's render() and it kept that class for the lifetime of that view.

In the new quilmes-master merged master, my view somehow loses the 'custom-view' class on the renderers update(), so now I have to again set 'custom-view' class on each update().

currently, it looks like the eventmachine native extension build fails during a run of gem install sproutcore. from a basic bit of poking around on google, it appears that eventmachine is a notorious pain to build on windows via gem.

I know this is really a problem with an upstream dependency, so it's more of a "heads up" than a "hey you guys fix this!"

Personally, I run sproutcore on either linux or a linux guest on windows, but I was curious what would happen in this case and wanted to share those results. I assume this is something that's changed in the build tools, as I recall that sproutcore 1.0 worked in windows without issue.

Using code from sproutcore-samples:

var pickerPane = SC.PickerPane.create({
  layout: { width: 300, height: 200 },
  contentView: SC.View.extend({
    layout: { top: 0, left: 0, bottom: 0, right: 0 },
  })
});
pickerPane.popup(itemView, SC.PICKER_POINTER, [3,0,1,2,2]);

This no longer shows the little pointer.

Screenshot here: http://i.imgur.com/g7wT7.png

Broken since commit e3bbb4a

Sean Eidemiller sent Charles Jolley a failing unit test

I have a Tab View that I am dynamically swapping the items that populate the Segment View portion depending on a selection in a Collection View.

Functionally everything is working as expected, but issue that I am having is related to "clipping" of the Segmented View portion. When selecting a Collection View item that would result in adding additional items to the Tab View, the Segmented View is "clipped" to only show an area based on the previous size.

If the browser window loses focus, all Segments seem to update to the desired rendered results. If an individual Segment is selected (even clicking the area that it should be showing in, even though it is not visible), that Segment is updated proper.

You can see a screen grab of the results at:

http://skitch.com/jaehess/dha6w/yesterday-149-messages

Chrome does seem to render everything properly, Safari is acting as described above.

SC.Button seems to only be used from SC.ButtonView, and as such, is only adding load-time overhead and making SproutCore more complicated and difficult to understand.

It should be folded in to SC.ButtonView.

It's not really clear what this does or if it's even necessary. If it's not necessary we should remove it. If it is necessary, we should document it better.

As you can see at http://localhost:4020/sproutcore/test_controls , SelectView's don't render properly.

The alert pane buttons have width and height 'auto', which becomes 0x0 px. I've verified this in the latest Safari and Firefox.

You can see a strange lightBlue bar next to Project and Task lists in Tasks app on iPad

When call on the DOM element corresponding to an instance of SC.TableCell, the $().view() does not return the SC.TableCell view. Instead it returns the SC.Table view.

The cause for this bug is that the function init() for SC.TableCell does not call sc_super(), and so SC.TableCell never gets registered in the SC.View.views object.

I found that sc_static('male.png') and sc_static('female.png') both translated to the path to the female.png image. I tried it with a longer filename like 'man.png' and 'thisisnotaman.png' and found the same result.

following the TableView tutorial using the master branch leads to:
javascript.js:6746TypeError: Result of expression 'this._viewRenderer' [undefined] is not an object.
in the renderViewSettings: function(context, firstTime) method

couldn't figure out where exactly it breaks but was able to reproduce several times
switching to the 1-4-stable branch avoids the issue

the table_view example on demo.sproutcore.com is broken as well

applyImmediately is an undocumented property in SC.TextFieldView. What does it do? How does it differ from continuouslyUpdatesValue?

According to the docs and code comments custom SC.ListView row heights should work something like this:

Pos.mainPage = SC.Page.design({
  mainPane: SC.MainPane.design({
    childViews: ['mainView'],

    mainView: SC.ListView.design({
      contentBinding: 'Pos.posController.arrangedObjects',

      contentIndexRowHeight: function(view, content, contentIndex) {
        return 36;
      },

      customRowHeightIndexes: function() {
        var content = this.get('content');
        var indexes = SC.IndexSet.create();

        if (content)
          indexes.add(0, content.get('length'));

        return indexes;
      }.property('[]'),
    }),
  }),
});

The console log shows customRowHeightIndexes is fired every time something is added to or removed from content but it only works the first time customRowHeightIndexes is fired. You can work around the problem by manually firing customRowHeightIndexesDidChange.

SC.SceneView is over a year old and doesn't take into account any of the new animation code. As soon as we get the animate method fully supported, we should switch it over to use that.

This can be seen happening in Tasks User Manager when you drag a user to change their role - works properly in Safari though

Currently we don't have any unit tests for the following SC.Store convenience methods:

• loadRecord
• loadRecords
• unloadRecord
• unloadRecords

SC.DataSource#retrieveRecords takes the parameters (store, storeKeys, ids), whereas SC.CascadeDataSource#retrieveRecords takes the parameters (store, storeKeys).

My proposed fix is to change all store.call(store, ...) with store.apply(store, arguments).
This is much safer, and prevents having to edit the code even if the parameters change.

EDIT:
I've added a pull request that fixes this and adds tests to SC.CascadeDataSource.
#91

Setting layoutDirection: SC.LAYOUT_HORIZONTAL on an SC.RadioView adds the proper class, but the styling is still vertical.

Peter Wagenet and I have both experienced occassional hints remaining in the field when entered and they become regular input text. I've experienced this issue on Chrome and Safari.

Showed this to Alex Iskander

Partial fix pull request sent

indexOf should always return -1 if no match. Returning false is very dangerous since it's easily confused with 0 which is a valid match response.

Encountered in 1.5

You can see this in the Tasks control bar at the bottom

Worked around in Tasks using hacks below

.mozilla.windows .control-bar .sc-radio-view.sc-layout-horizontal .sc-radio-button {
margin-top: -12px;
}
.chrome.windows .control-bar .sc-radio-view.sc-layout-horizontal .sc-radio-button,
.safari.windows .control-bar .sc-radio-view.sc-layout-horizontal .sc-radio-button,
.safari.mac .control-bar .sc-radio-view.sc-layout-horizontal .sc-radio-button {
margin-top: 4px;
}

SC.TextFieldView uses a function called getValidatedValueFromFieldValue that is neither implemented nor documented anywhere in the SC codebase, causing it to throw an exception.

See #10

Issue was present before I upgraded jQuery to 1.4.3 and after.

If you have two views side by side with mouseEntered/mouseExited events and mouse into the second view from the first, the second view's mouseEntered is called BEFORE the first view's mouseExited.

This causes a lot of headache if you want to e.g. have a shared state between a set of list items.

It's also counter-intuitive when you have a knowledge of statecharts.

It seems there is a problem with SC.RecordArray.refresh(), or rather SC.Store.refreshQuery().
After adding and removing records from the back-end, calling refresh() adds the new records, but doesn't remove deleted ones, so the resulting SC.RecordArray doesn't match the result of the fetch query.

Calling reset() on the stores doesn't help (or doesn't work?, they still contain the deleted elements), but will cause an internal inconsistency.

If you login as 'SG' no password to http://tasks-demo.appspot.com and click on the icon of any red task you will get slide-out task editor. Click on the watches button at the bottom of that editor and you will see the problem. This happens only on SC master (quilmes).

If you use MasterDetailView and use its popover feature, the popover will be openable when in portrait mode. However, if you open it, then switch to landscape (with it still open; it closes on its own), and then switch back to portrait and try to open it again, it breaks.

The bug may also be there if you close it yourself, switch to portrait, and then switch back.

after switching from 1.0.46 to 1.4.1 I see this warning:

WARN ~ Could not find entry 'system/loader' required in sproutcore/bootstrap:source/setup_body_class_names.js

and then none of my sc_requires are found, the files are loaded anyways but not in the right order which breaks the app.

turned out to be the sc_require parser not liking camel case anymore (used to in 1.0)

Showed to Alex - regression introduced by quilmes

In SC 1.0 and beyond, seems that SelectView has had a bug that triggered the following error message:

TypeError: Object [object DOMWindow] has no method 'parentMenu'

This error still ails SC 1.4.2. A fix was proposed in http://goo.gl/cjSN

An SC.Record with an SC.DateTime attribute is not parsing JSON datasources where the field is returned as a Number.

There are some problems working with nested records:

1. When the nested record is first created, it is added to the store's changelog, which would be fine except that the store is not being updated when the nested record's status changes from READY_NEW to READY_CLEAN. Therefore, calling commitRecords() on the store causes all of your nested records to be duplicated. To see this, use store.readStatus() on the nested record, if you use nestedRecord.get('status'), then the store's status will be updated.
2. Modifying a nested record within a nested store does not work. The nested record's _parentRecord does not get transferred over into the nested store version and therefore, the nested store parent record is never notified of any changes. Committing changes from the nested store ends up with only the nested record's status being READY_DIRTY, but the changes in the nested record don't exist in its parent record and therefore are very easily overwritten whenever the parent record updates.

Hi,

IE8 will not run the latest Quilmes.

To reproduce:

Using the latest sproutcore gem (1.4 pre)

• sc-init IETest

  cd ie_test

  mkdir frameworks

  cd frameworks

  git clone http://github.com/sproutcore/sproutcore.git

  cd sproutcore

  git checkout quilmes

  cd ../..

  sc-server -v

• open IE8 in a Parallels instance, visit the sc-server site (in my case, http://rklancer-unibody-2.local:4020/ )

• wait for the Welcome app to come up

• Click around; nothing happens.

• Open the Developer toolbar (Tools -> Developer Tools)

• Click the 'Console' tab if it's not already selected. You should see several 'Invalid Argument' errors referencing this line from jquery.js:

  if ( set ) {
      style[ name ] = value;
  }
  

The issue appears to be that view.js is setting the 'zIndex' and 'backgroundPosition' style attributes to null using jquery, and IE8 does not tolerate a null value of these attributes

UPDATED: I fixed the issue with a "works for me" in commit 5c87a88 in branch 'better-quilmes-ie-fix' of my fork but need guidance regarding:

1. Is this really a jquery issue? Should jquery normalize these 2 attributes?
2. Does my change cause any other tests to fail?
3. Should I/how can I create a failing test to demonstrate the problem?

You can see the problem on http://tasks-demo.sproutcore.com login 'SG' no password
Goto Actions (gear) menu and choose Export - you will see scrollbar missing - two finger scrolling still works though

If you have a ScrollView within a scene of ContainerView or SceneView, when you transition back to that scene the previous scroll location is not re-set.

This issue became apparent when I attempted to run the build tools from an ant task. I believe the java VM defaults to ascii encoding and so I was get a Fatal encoding error. Wycats helped me out in irc, and i discovered "Encoding.default_external" was ASCII while running from ant but UTF-8 when running from the command line.

As I quick fix to the problem, I have added "Encoding.default_external="utf-8" to the top of my sproutcore.rb to ensure it is always used.