As the summary says the whole WML loading / page cache / history handling is flawed.
It depends on several hacks in FrameLoader / WMLGoElement / WMLCardElement / WMLPageState.
* There should be no need to track the history length on our own
* <go> / <a> / <anchor> / <do> / <prev> shouldn't need to deal with the FrameLoader directly (currently setting setForceReloadWmlDeck()
* FrameLoader needs to detect WML documents and handle them approriate (always reload, no caching, etc.) The tricky part is detecting our layout tests which embed WML content in a XHTML document (which is really 'hacky', but needed for our automatic testing facilities).
Fixing these issues should stabilize the loading area between running the layout tests standalone in DRT and using WML docs directly.
Uploading a patch with new testcases soon.
Created attachment 33553 [details]
Landed in r46418.