Draft 'install alongside Windows' test case

[Adam Williamson]

On Thu, 2011-09-01 at 05:49 -0400, Al Dunsmuir wrote:
> On Wednesday, August 31, 2011, 10:27:27 PM, Adam Williamson wrote:
> > On Wed, 2011-08-31 at 20:45 -0400, Al Dunsmuir wrote:
> >> On Wednesday, August 31, 2011, 8:08:19 PM, Adam Williamson wrote:
> >> > Hey, folks. I just threw together a quick draft of an 'install alongside
> >> > Windows' test case - we have this as a final criterion, but no test case
> >> > for it as of yet. Here's the draft:
> >> 
> >> > https://fedoraproject.org/wiki/User:Adamwill/Draft_QA_Testcase_install_alongside_Windows
> >> 
> >> > any comments, questions, suggestions welcome! thanks.
> >> 
> >> I  think that you need to be specific as to what you mean by "Windows"
> >> - especially given the different behaviours with GPT and MBR.
> 
> > Well, no version of Windows yet sets up a GPT disk label, and anaconda
> > is supposed to leave existing MSDOS disk labels around.
> 
> True,  but  anaconda  does  let one create additional empty partitions
> that  could  be  used to install Windows after Fedora (creating a dual
> boot  system  in the opposite order).

The release criterion explicitly doesn't support this; installing
Windows after Linux is known to be Doing It Rong, or at least, making it
harder than it needs to be.

> Where  a  Windows variant does support GPT, it would be useful to know
> if  it  tolerates  installing in a partition created during the Fedora
> install.

Useful, sure. Not really part of our criteria or something I'm terribly
worried about for release validation testing, though. I mean, if you
want to write something up to be added to the matrix as an optional test
or something, go for it...but I don't think it's something we need to
cover in just a simple basic test case to validate the criterion.

> >> There need to be tests (with release blocking on fails for):
> >> * XP 32-bit (both FAT and NTFS)
> >> * Vista (32 & 64-bit)
> >> * Windows 7 (32  &  64-bit)
> >> * Windows 8 preview (32 & 64-bit)
> 
> > Remember, the tests go into a matrix which has both 32-bit and 64-bit
> > columns, so there's no need to have different tests.
> 
> I'm not talking Fedora 32-bit vs 64-bit, I'm talking Windows 32 vs 64.

The results page doesn't say 'Fedora' either =)

> Making  sure  one  records  the  Windows  variant separately minimizes
> confusion,  and  allows  tests  by  multiple  reporters  to  be merged
> together reasonably.

Again, that's not really something that goes in the test case itself,
more in the results table. I don't really want to make that any longer
than it *needs* to be. If we do get lots of people testing, and
confusing results, we can lengthen it appropriately.

> > I'm not sure it's realistic to list separate test cases for every
> > version of Windows known to man; I left it generic on purpose so that
> > people can test with whatever Windows they happen to own. The version of
> > Windows that's present shouldn't ever make an awful lot of difference,
> > anyway, since Fedora doesn't have to *do* anything with it besides
> > identify it. Ditto FAT vs. NTFS: we're intentionally *not* covering
> > partition resizing here, as it's not supported. About the only issue
> > would be ensuring os-prober can recognize all Windows variants.
> 
> The  grub  utility  that  does  probing  needs  to be tested against a
> variety  of Windows variants.  Accurately recording which variant that
> was tested against was the point of my post.

Well, my take on that is that if someone actually hits a failure,
they'll need to file a bug report, and that information can be included
in the bug.

> >> I  would  suggest  that  you  also  want to test against a DOS variant
> >> (especially FreeDOS since it is FOS).
> 
> > This isn't in the scope of the release criteria, and I don't think it
> > really needs to be. We're concerned with the common case of 'I want to
> > install Fedora alongside Windows'. Installing alongside DOS is pretty
> > corner case-y these days.
> 
> FreeDOS  documents  grub  booting. Again, the grub _probing_ should do
> the  right  thing, or at least do no harm.

Should? Sure. But I'm concerned with release validation, here, and
that's not part of it. I'm not trying to write test cases to cover every
possible function of os-prober here, just a test case to give a basic
procedure to validate the release criterion we have in place, which is
specific to Windows.
-- 
Adam Williamson
Fedora QA Community Monkey
IRC: adamw | Twitter: AdamW_Fedora | identi.ca: adamwfedora
http://www.happyassassin.net