The 'installation source' spoke is meant to only show the option to install from an ISO file if there's a viable source for one. To put it in negative terms, if the currently-planned partitioning scheme involves reformatting all available disks, the option is supposed not to be shown because there's no place you could possibly get an ISO file from. This works OK if you only enter the 'installation source' spoke after going to 'installation destination' and picking such a partition scheme - say there's only one disk on the system, and you pick it for auto-partitioning in F18 Alpha (which means it'll be reformatted). If you then go to 'installation destination', the ISO file option is not shown. However, it breaks if you enter 'installation source' *before* entering 'installation destination'. The ISO file will be shown when you first enter 'installation source' - which is correct - but it'll still be shown even if you then go into 'installation destination', select all available disks for formatting, and then re-enter 'installation source', which is *not* correct. In this case, anaconda should recognize that the planned partition scheme has changed, and no longer show the option. (It should also probably warn the user and void the 'installation source' selection if the user picks an ISO file as the install source and then selects the partition where the ISO file resides for formatting in 'installation destination'; I haven't checked yet if that's in place). dlehman reckons this is because "we have code to set it visible on refresh, but no code to set it invisible on refresh (if I'm seeing this right)" and suggests http://fpaste.org/B2Ch/ as a possible patch.
That looks like a suspiciously short patch, but I could put together an updates image to test if you have a test setup. I hate testing hdiso.
You don't actually have to complete an install or even have an ISO image on the test system to test this bug. You just have to clicky clicky between the spokes as described in the report. I'm just testing it in my usual test VM. I'll happily test it if you give me an updates image.
I don't think that patch works.
But now I have a better plan - one that excludes the blob's stupid mother.
Confirmed fixed now (Beta RC1, but certainly much earlier) - cannot reproduce as per above.