While looking at the GXT Demos, I noticed that on Basic Grid, if I resize the grid and refresh the browser, the grid breaks, as shown in the screenshot below. If I resize the grid again (manually), everything becomes normal again. Is this a bug?
This happens quite frequently on Chrome, but intermittently. I had version 23x at that time. I observed similar issue when I recreated that example grid (with toolbar) in my test application. The toolbar was broken in my app occasionally.
I just do a normal browser window refresh. I didn't clear the cache.
Oh BTW, I saw a similar issue in my very first HelloGXT app as well where the button text would not show up and instead it would show >> with a drop down button (just like you see in the screenshot in my OP). Upon clicking the drop down button, it would show the original button text.
I'll post screenshots if I see it again in other widgets.
Thanks for the updates - I've managed to see this very intermittently in webkit only (Chrome and Safari), and it *seems* to be related to network connection speed somehow: the slower the connection, the more frequently it seems to occur. That said, order of rendering components or loading images shouldn't be affected by this, since all code runs synchronously anyway, and all images (in Chrome anyway) should be inlined in the main .cache.html file by ClientBundle.
It seems very unlikely that IE- and Chrome- specific issues have the same root cause, but it is possible - thanks for the reference.
There is substantial logging disabled by default within GXT that describes hard to debug issues like layouts and focus operations. If the java.util.Logging tools are already enabled, you can ask GXT to output these additional messages by adding this line to your .gwt.xml file:
This block is already present in the explorer, but the extra GXT logging is not enabled to keep the console spam to a minimum.
I've filed this internally to try to track it and correlate other discoveries with it, but as it is intermittent, we're still at the data-collecting stage. We would welcome any other details about projects that experience this, especially any that can reproduce it more often than not, to allow us to add even more logging until we can point to the root cause.