On Sat, Dec 5, 2020 at 4:19 PM Matthew Miller <mattdm(a)fedoraproject.org> wrote:
On Sat, Dec 05, 2020 at 08:32:28PM +0000, Naheem Zaffar wrote:
> I think we need more than just better instructions here.
>
> If you use automatic partitioning, there is also a "leap of faith" step
> where you have to hope that the installer does the right thing, but without
> any way of verifying it.
>
> If you don't use automatic partitioning, it is because you already know
> enough to get past this stage.
Serious question since I don't dual boot -- can the automatic partitioning
handle the Windows 10 dual boot situation? Should I have just told the user
to use that?
It does handle it including shrinking the Windows primary volume
(NTFS), and reusing the existing ESP.
> In my opinion the most ideal scenario here would be a Windows
based
> pre-installer - it runs the necessary checks that Fedora can be installed,
> sets up partitions and then reboot to finish installation.
Sounds interesting -- how would this help?
It'd make shrink Windows' problem rather than Anaconda's. It does do
an online shrink, creates unallocated space which Anaconda sees as
"free space" to just install to without bringing up the Automatic UI's
"reclaim space" dialog.
The reclaim space dialog kinda depends on the user choosing the
correct partition to shrink. It should be obvious which one it is (the
big one). This UI is somewhat more intuitive, with a slider for
choosing the shrink amount/ratio. The Windows UI is a wizard style
interface and presents a bunch of numbers and jargon that's not
entirely obvious whether I'm specifying size for the unallocated space
(yes, it is) or the Windows partition.
In about 10 years, I can recall one bug, which haha - I ran into. And
it did break the file system. I forget if it was repairable, mainly
because it was an expendable test subject.
There is a release criterion for dual boot installation of Windows. It
has to work for both automatic and custom partitioning UIs, including
successfully shrinking. There's also a test case for it. I don't know
if it's tested in openqa though.
--
Chris Murphy