Re: resizing with gravity

Russell Shaw wrote:

Sasha Vasko wrote:

Russell Shaw wrote:

Sasha Vasko wrote:

An X app might open a grid of top-level windows, and rearrange them
as more windows are added or removed.

No application should ever rely on being able to do that.

No app for use by general users should expect that. However, for a
developer making a program for a specific platform can do anything
they like, using any assumptions they know to be true.

Yes and one of this assumption is that app has no control over its position on the screen. On X Window platform that is.

Such apps may exist primarily for embedded systems without a window
manager, but it is still useful to be able to use those apps where
a WM is running. In that case, the WM must have a policy setting that
allows the app to run as designed, or in a semi-managed way.

Yes, that is what I'm saying - user (via window manager configuration) determines how windows are placed. Not the client application.

In fact I suggest that we explicitely state the opposite: "Clients MUST NEVER EVER include x and y in ConfigureRequests" Although most smart and responsibly implemented clients already know that, and those who suffer from delusion of grandeur, will always ignore such statements anyways, so really there is no point in amending specs.

Clients should only be able to change size of the top level windows and then let window managers reposition them in accordance to its policies. Which is were gravity comes into play ( from WM_NORMAL_HINTS ). And again, this gravity should be choosen by clients based on user request of initial geometry, and should not be mangled by the client itself.

Window managers have no place in dictating how an application manages
its top-level windows.

Last time I checked Window Managers are REQUIRED to intercept ConfigureRequest. It is my humble guess that WM does have a responcibility of placing windows as it sees fit.

It is the responsability of the window manager to place windows as the *user*
has specified it should be done.

That is correct: *user*, but not a client app.

The user may tell the WM to not modify the actions of one app that they
may be testing for use on a platform that has no WM.

If you are testing an app for platform without WM - then run it in server without WM

Window managers exist solely for the purpose of the *user* to specify
how top-level windows should be managed.

Yes and that is why window managers have THEIR OWN CONFIGURATION settings. And this is why there are so many window managers available! But after user has selected particular window manager and configured it how it wants windows to be placed - client have no business whatsoever interfering with that!

Clients will do whatever they have been designed to do. A few clients
may be designed to function well with or without any WM, so when a WM
is running, it will modify the clients actions according to the users

Its not difficult at all for a client to recognize that there is no WM running and take over responsibilities of its own placement in such case. Somehow I don't see any noticeable number of applications worreing about running without WM though.

If a user wants to disable automatic window management for one app, then
so be it. A WM that doesn't allow this is broken.

ConfigureRequest is still handled by WM anyway you go about it. Clients CANNOT possible determine appropriate x and y as they could never precisely determine the size of the decorations at the time of request. There may not even be a decoration! There could not even be such thing as free placement of the window - WM may tile them all together, or use tabs, and allow only one window on screen at a time. There could be all kinds of placement policies out there, and no client will ever be able to account for all of them.

For a client that is designed to run multiple top-level windows even with
no WM running, the client just needs to do its measurements on the root
window (which could be just inside a WM frame, if a WM is running).

Note that this is for clients specifically designed for WM-less operation,
which are very few, but not non-existant. A developer will need to emulate
WM-less operation in their WM to test or use these apps on their desktop system.

Well telling WM to pretend that does not exist is an idiotic way of testing such apps - you can never be sure that it really does that. Run nested X server if you really need to test that, or run a second head without a WM. Whats so complicated about it ?

Simple example: a terminal application.
You start it like so: xterm -geometry 100x25-0-0

Terminal then gets placed at the bottom-right corner of the screen. Now what would happen if user decided to change font size in it ? Do you really want xterm to try and calculate what its x and y should be for ConfigureRequest??? I don't. And I doubt that xterm developers differ with me on that one.

If an application was designed with the assumption that a WM is always
running, then the offset calculation isn't required.

If the app was designed assuming a WM may not be present for a significant time, then i *would* expect the app to do the offset calculation.

Then this must not be part of Extended WINDOW MANAGEMENT specs, if your goal is to account for when WM is not running.

It doesn't need to be a part of the Extended WINDOW MANAGEMENT specs, because it is just a matter of the user setting their WM to switch off or modify handling
of a certain client they wish to.

The point is that some clients *do* need to set their size and position on the root window (as opposed to clients having no business setting their position). When a WM is running, the positioning can be ignored if desired, by the users
WM settings.

If its ignored then its not taking effect - ConfigureRequest just gets discarded. But if it is not discarded, and honored, any but the most primitive of window managers will have to take into consideration all kinds of different thing, such as AvoidCover windows, Struts, limits of desktop area available, virtual desktop size, tiling with other windows, decorations, all kinds of garbadge like that, end result of which is that final position is almost always different from what client has requested.

Sasha Vasko

[Date Prev][Date Next]   [Thread Prev][Thread Next]   [Thread Index] [Date Index] [Author Index]