Go the URL above, and delight in the wonderful things it brings. Now close all the windows, leaving a small one until last. Now open a new browser window. Note that it comes up small, taking its size from the last window closed (which was a popup window). This is wrong; we should never store size data for popup windows, because the user has not specified their size.

reassign to waterson since there needs to be a way to say don't make this window persistant. I am not reallyw to tell something is a popup either but I'll worry about that after we have a way to disable persistance.

davidm: _nothing_ persists by default. you need to explicitly tell XUL to persist attributes (see the "persist='...'" in navigator.xul). so i'm pretty sure that this is a problem with the core persistence implementation. cc'ing law, who did the original navigator.xul persistence stuff.

Sorry I should have been clearer. A popup window is a browser window. Nothing more, nothing less. So for some browser windows we need persistance. For others we don't.

right, but the persistence stuff is added for _every_ navigator.xul window, because "persist='width height x y'" is set on the 'window' tag in that file. You don't need to do that. You can decide which windows you should persist the size information for, and for those windows, call document.persist() with the ID of the window tag and the name of the attribute you want to persist. Do this in your "onunload" handler. Sorry about the lack of documentation. For now, see news://news.mozilla.org/37CF8F34.BAC4059D%40netscape.com

I'm closing this as a dup of bug #13208. There's some discussion there, but my current theory is that we need to distinguish windows opened by the user from those opened by spam-spewing web pages (no offense). My plan was to add a property to browser windows that indicates whether it should save the window size when it is closed. This property would be false by default and set from navigator.js code that opened new windows. The only thing that stopped me from implementing this fix immediately was that the initial browser window is opened from C++ and I didn't know how to set this property from there (I'm sure it's possible one way or another). So I don't think this requires some major architectural/design work, just a little bit of code. BTW, if one were to do something like: var newWindow = window.openDialog( "chrome://navigator/content/", ... ); newWindow.document.getAttributeById( "main-window" ).removeAttribute( "persist" ); would that automatically prevent changes to persistent attributes on that element from being persistent without affecting other instances of the same xul document? If that works, then this simplifies things because we can thus use the DOM apis to change the "property" of the window (from C++). *** This bug has been marked as a duplicate of 13208 ***

Yes, removing the 'persist' attribute should work like you'd expect. If it doesn't open a bug on me.

Looks like a dup. Marking VERIFIED.