Alok,
Alok Aggarwal wrote:
> Hi William,
>
> On Thu, 9 Apr 2009, William Schumann wrote:
>
>> There are two problems addressed by this bug.
>>
>> 1) After initial AI installation, the swap slice is regarded as user
>> data in subsequent AI installations instead of scratch
>> 2) The swap slice space allocation in TI is not done with any
>> awareness of AI customization
>>
>> The fix involving making AI aware of 1 and for AI to perform 2
>> instead of leaving everything to TI, which allocates the swap space
>> for GUI.
>
> So, from a high level, is it fair to say that the policy with respect
> to preserving slices by default isn't being
> changed?
Yes, the policy for preserving slices by default does not change.
> And, the change you're making here is to create
> a swap slice if either s1 doesn't exist or s1 exists with
> a V_SWAP tag. Correct?
The decision of whether a swap slice is created depends upon the memory
size of the machine.
>
> Also now that 7718 has been reversed in the source code
> only machines with less than 700MB memory will see this
> problem on every other AI install after the *first* install.
I think it would fail *all* AI installs after the first one (that didn't
explicitly delete slice 1 per the workaround); otherwise, what you said
is true. Also, without the fix, problem 2) above still exists - if a
swap slice is called for, any slice customization or any existing slices
which should be preserved will be destroyed instead.
William
>
> Given this, the question in my mind is whether it is worth
> making complicated changes like these this late in the
> release for a relatively small set of machine. Or, to think
> through the problem we're solving more carefully and defer
> it to the next release.
>
> Do others have an opinion on this?
>
> Alok
