70380 - nsISingleSignonPrompt needs to be pushed down a level, making it private.

Status: VERIFIED FIXED

(Core Graveyard :: Embedding: APIs, defect)

Description

[Judson Valeski]

see URL for more info (the notes on nsIPrompt also talk about
nsISingleSignOnPrompt).

Comment 1

[Conrad Carlen (not reading bugmail)]

targeting 0.8.1

Comment 2

[Conrad Carlen (not reading bugmail)]

-> mozilla0.9

Comment 3

[Conrad Carlen (not reading bugmail)]

Attachment: tree-wide diffs

Comment 4

[Conrad Carlen (not reading bugmail)]

Sorry about the non-contextual diffs. Good news is that MacCVS can now generate
diffs for a whole tree, bad news is that it does let you pick diff options so
you get normal diffs. Contextual diffs will follow.
The patch splits the username and password prompting methods of nsIPrompt into a
new iface nsIAuthPrompt. Those methods (prompt, promptPassword, and
promptUsernameAndPassword) which remain on nsIPrompt are specialized for
password UI as opposed to password persistence. With this specialization,
single-signon can now call the password prompting methods on nsIPrompt instead
of the dreaded UniversalDialog.
The only prompt iface which the embeddor will ever see is nsIPrompt. They will
never need to be concerned with password persistence. The persistence layer is
hooked up to their nsIPrompt under the covers in the webbrowser's docshell tree
owner.
Jud, Rick - Can you review and super-review?

Comment 5

[Conrad Carlen (not reading bugmail)]

Attachment: tree-wide context diffs + embed samples

Comment 6

[Conrad Carlen (not reading bugmail)]

What the patch does:
(1) The prompting methods for which it was possible to specify a URL domain and
whether and how to save the password have been moved into a new iface -
nsIAuthPrompt.
(2) Since nsIPrompt now handles only UI instead of UI and persistence, it's more
capable. The prompting methods which are now left on nsIPrompt allow for names
and passwords to be in/out params and to have an optional checkbox. This means
UniversalDialog does not have to be used for what should be simple dialogs.
(3) nsISingleSignonPrompt is completely from embeddors. It gets hooked up to
their nsIPrompt automatically from the docshell tree owner which is below the
level they can see.
(4) A new component (which is pysically part of the embedComponents DLL) is
introduced for the case of nsISingleSignonPrompt not being available
(installed). It bypasses persistence by mapping nsIAuthPrompt methods to nsIPrompt.

Comment 7

[Conrad Carlen (not reading bugmail)]

Whoops, the line above which says:
> (3) nsISingleSignonPrompt is completely from embeddors.
should say:
(3) nsISingleSignonPrompt is completely *hidden* from embeddors.

Comment 8

[Stephen P. Morse]

Conrad, thanks for copying me on this bug.  Prior to today I wasn't aware of it 
and thought that the nsiPrompt work would have to be done as part of bug 70382.

Comment 9

[Judson Valeski]

r=valeski on the 03/26/01 15:07 patch, after the full check for password pointer
validity in nsBrowserWindow::PromptUsernamePassword() is made. We'll need to
cover any commercial instances as well before checking in though.

Comment 10

[rpotts (gone)]

sr=rpotts...

Looks good - it sure is a big patch :-)

-- rick

Comment 11

[Stephen P. Morse]

Conrad, be forewarned that I just checked in my patch for bug 70382.  As a 
result, you'll need to update the following files mentioned in your patch:

   nsIWalletService.idl
   nsWalletService.cpp
   singsign.cpp

Comment 12

[Conrad Carlen (not reading bugmail)]

Steve, next time I merge my changes with the tip (and certainly will before
checking in), that'll be covered along with any other changes.

Comment 13

[Conrad Carlen (not reading bugmail)]

Fix checked in.

Comment 14

[Michael Dunn]

Correction: Changing QA contact for the Embed API bugs to David Epstein.

Comment 15

[David Epstein]

reassign qa contact to Dharma.

Comment 16

[dsirnapalli]

-- verified.