Hi Sir,
I am investigating an intermittent dialog visibility issue on Linux with:
* Qt 5.12, using the XCB platform plugin
* TWM 1.0.9
* X.Org server 1.20.1
Occasionally, after an action that should display a modal dialog, the main
window becomes dimmed but the dialog is not visible. The application remains
responsive, and ongoing display updates continue.
Keyboard input or interaction with another UI control has reportedly made the
dialog appear. However, we do not yet have a complete failure-and-recovery
protocol trace, so these observations do not establish whether mapping,
stacking, focus, or application behavior is responsible.
We reviewed the application code involved in this dialog-display sequence.
Within the reviewed path, we found no explicit call to hide the dialog or unmap
its native window while it is expected to be displayed.
In one failure-time window-tree capture, the expected dialog XID appeared as a
direct child of the root without a TWM frame. Unfortunately, that capture did
not record its map state or override-redirect attribute. Therefore, it does not
establish whether the window was unmapped, withdrawn, or otherwise unmanaged at
that moment.
Successful traces show the expected sequence: client MapWindow, TWM MapRequest
handling, reparenting, and mapping of the client and frame. Those traces do not
explain the intermittent failure.
The issue occurs very intermittently, and enabling X protocol tracing makes it
significantly harder to reproduce, possibly because tracing alters the timing.
We would appreciate guidance on:
1.
Which failure-time attributes and protocol events best distinguish:
* A client that has not requested mapping
* A redirected MapRequest that the window manager has not yet handled
* A mapped client with an unmapped frame or ancestor
* A viewable window hidden by stacking
2.
Whether TWM has event-handling or interactive paths that can temporarily delay
MapRequest processing, and how best to identify them.
3.
How to trace client map/unmap requests and TWM handling reliably across
separate X connections.
4.
Whether there are known issues relevant to this combination of versions.
The current plan is to capture client/frame map states, parent relationships,
override-redirect, WM_STATE, WM_TRANSIENT_FOR, stacking order, input focus, and
correlated client/TWM event traces during the failure.
The root cause is not established, and we are not where is issue .
Recommendations for minimally intrusive diagnostics would be particularly
helpful.
Thanks and Regards;
Nilesh
