[toolkit] Empty PIN/password/passphrase field appears to contain 4 characters

[kevin gunn]

minor interrelated issues & probably needs design input
1) visual inconsistency of the pin text box prompt shows 4 "bullets" implying needing 4 characters, but the passphrase is text box is empty
2) selecting the pin text box prompt, the "bullets" remain, and don't respond to delete, they don't disappear once typing begins (i would think dissappear on text box selection)
3) having a "show password" would be nice

------------
Desired resolution: Passcode/password fields should never contain bullets when you haven't entered anything in them.

[Michał Sawicz (saviq)]

It's the hint that you can see, bug in SDK that they don't disappear.

[Matthew Paul Thomas (mpt)]

This is not only a problem in the first-run setup, but in every password field, for example the "Switch to Passcode"/"Switch to Passphrase" dialog.

[Matthew Paul Thomas (mpt)]

Fields should never contain bullets as hints. Hints are a bad idea in general, but bullets as hints aren't even telling you anything. Please just remove them.

[Sebastien Bacher (seb128)]

that's a duplicate of bug #1352953, not sure if it should be marked this way though since that one has design recommendations

[Launchpad Janitor (janitor)]

Status changed to 'Confirmed' because the bug affects multiple users.

[John McAleely (john.mcaleely)]

This has been recently customer reported in bug #1403878

[Pat McGowan (pat-mcgowan)]

This is a bug in the dialog's use of textfield. We currently set an inputMask of "9999" which causes the pre-population of asterisks. The correct thing to do is set maximumLength to 4 and not set an inputMask. The experience is that the field stops taking input after the 4th digit, and properly displays the digits one at a time as they are input to a blank field.

[Iain Lane (laney)]

The point of the input mask is to make the field only accept digits.

[Ken VanDine (ken-vandine)]

I agree with laney, the inputMask is needed. I think this is probably an SDK issue.

[Pat McGowan (pat-mcgowan)]

There is no way to enter other than digits already, the field has inputMethodHints set to Qt.ImhDigitsOnly and the OSK is shown in digits mode,

[Ken VanDine (ken-vandine)]

That only works around the issue, and will work properly in the case of the OSK. But on a desktop, or a device with a keyboard attached, it won't work.

[Pat McGowan (pat-mcgowan)]

fwiw this seems to work everywhere although may be better solutions

            TextField {
                id: currentInput
                echoMode: TextInput.Password
                inputMethodHints: Qt.ImhNoAutoUppercase |
                                  Qt.ImhSensitiveData |
                                  Qt.ImhDigitsOnly

                maximumLength: 4
                validator: IntValidator{bottom: 0; top: 9999;}
                onTextChanged: {
                        confirmButton.enabled = text.length == 4
                }

            }

[Ken VanDine (ken-vandine)]

A slight improvement, instead of changing confirmButton.enabled with onTextChanged, you can use the acceptableInput property to control enabled for confirmButton.

    enabled: currentInput.acceptableInput

[Jonas G. Drange (jonas-drange)]

<email address hidden> wrote:

> A slight improvement, instead of changing confirmButton.enabled with
> onTextChanged, you can use the acceptableInput property to control
> enabled for confirmButton.
>
Do we get a chance to restrict passcode input to exactly four digits, then?
The input mask did this for us, while the IntValidator does not.
​

[Jonas G. Drange (jonas-drange)]

Maybe we'll just change bottom to 1000.

[Pat McGowan (pat-mcgowan)]

acceptableInput doesn't ensure 4 digits which is why I used the length check, and there is no way to specify minimum length.
Using 1000 as a bottom would not allow a 0 in the first digit so do not think that works.

[Pat McGowan (pat-mcgowan)]

acceptableInput would work if the validator was extended to include a length check

[Jonas G. Drange (jonas-drange)]

How though? There is no Validator qml element, nor does IntValidator expose
anything of interest [1].

[1] http://doc.qt.io/qt-5/qml-qtquick-intvalidator-members.html

[Pat McGowan (pat-mcgowan)]

We can define a custom validator, not sure if there is also a way to combine validators

[Pat McGowan (pat-mcgowan)]

Here's the better way

                validator: RegExpValidator { regExp: /\d{4}/ }
                onTextChanged: confirmButton.enabled = acceptableInput

[Ken VanDine (ken-vandine)]

I'm reopening the uitk task for this. We've worked around this in system-settings, but I think we should handle this in the toolkit so we get consistent behavior across the platform.

[Pat McGowan (pat-mcgowan)]

@ken are you suggesting a shared "passcode" component?
Where else are we presenting this? Login, wizard, ...

[Iain Lane (laney)]

The correct solution is to not have the bullets displayed for a TextField in TextInput.Password mode using an inputMask. No placeholders are displayed when echoMode is TextInput.Normal.

See: https://code.launchpad.net/~jonas-drange/ubuntu-system-settings/security-pass-input-fix-1357548/+merge/245980/comments/607924

I don't see any analysis of the work required to have a fix in the SDK, instead jumping straight to a workaround in u-s-s. Did this happen somewhere else?

[Ken VanDine (ken-vandine)]

@laney the sdk component behaves the same was TextField does in Qt, it's currently as designed (afaict). I think this is more of a design issue, if we don't want to display those for passcodes then we should handle that in the SDK. So we get a consistent experience.

[Launchpad Janitor (janitor)]

This bug was fixed in the package ubuntu-system-settings - 0.3+15.04.20150112-0ubuntu1

---------------
ubuntu-system-settings (0.3+15.04.20150112-0ubuntu1) vivid; urgency=low

  [ jonas-drange ]
  * [security] replace inputmask for validator and maximum length (LP:
    #1357548)
 -- Ubuntu daily release <email address hidden> Mon, 12 Jan 2015 16:43:30 +0000

[Launchpad Janitor (janitor)]

This bug was fixed in the package ubuntu-system-settings - 0.3+15.04.20150112~rtm-0ubuntu1

---------------
ubuntu-system-settings (0.3+15.04.20150112~rtm-0ubuntu1) 14.09; urgency=low

  [ Ken VanDine ]
  * [security] replace inputmask for validator and maximum length (LP:
    #1357548)
 -- Ubuntu daily release <email address hidden> Mon, 12 Jan 2015 21:51:55 +0000

[Noemí (noemi-gallego)]

This is still a bug in R7 Version. Could you tell me more or less in which version this is going fixed?

[Jonas G. Drange (jonas-drange)]

@noemi-gallego this landed in RTM 14-Jan-2015 09:40 [1] so I assume the next RTM image produced after that date.

[1] http://people.canonical.com/~lzemczak/landing-team/ubuntu-rtm/192.commitlog

[Matthew Paul Thomas (mpt)]

Verified in Ubuntu 15.04 devel-proposed r78.

[Zsombor Egri (zsombi)]

To handle this in SDK we would need to have a separate PasswordField component, which can have the design requested functionality in case that is different from the one provided by Qt TextInput.

[Matthew Paul Thomas (mpt)]

Zsombor, a separate PasswordField component would not fix the bug at all. On the contrary, it would make the situation even worse. There would now be two ways of creating a password field rather than one, for no benefit, making the SDK harder for developers to learn and understand. And TextInput.Password would *still* do the wrong thing every time you use an inputMask.