msi: begin support for finding network last sources in find_published_source

[Aric Stewart]

Hello,

   No i never saw the e-mail that you send to Jeremy.

   I will admit that i did not dig really deep for this patch as it 
appeared to me that the fix was obvious (there was even a fixme in the 
code saying we needed to do what the installer was asking for)

   There are about 4 CAB files that the service pack asks for to install 
(and only installs a few files out of each) that are accessed by this 
method.  Each one does not exist on the original media at all and only 
seems accessable via the LastSource location.

   I am curious what tests you did to prove that you did not need the 
original media or even the copied cab files?  Did you do a test on 
windows where you deleted them all and the upgrade worked?

   I will investigate some more, but i do not feel that this patch is 
incorrect, even if it should not be needed. Am I right there?

thanks,
-aric

James Hawkins wrote:
> On Thu, Feb 19, 2009 at 12:28 PM, Aric Stewart <aric at codeweavers.com> wrote:
>> Fix for office 2007 sp 1 install.
>> ---
>>  dlls/msi/media.c |   33 +++++++++++++++++++++++++++++++++
>>  1 files changed, 33 insertions(+), 0 deletions(-)
>>
> 
> I believe you're going down the wrong road with respect to the Office
> 2007 SP1 installer.  None of the files installed/patched by SP1 need
> access to the original media, or stored local copies of the cab
> archives.  With the fullfile install (which I'm not sure there's
> another version), you should be able to remove the original media,
> erase the stored cab files, and successfully install SP1.  The fact
> that our implementation tries to access those cabs or the original
> media is the bug.  There are many conditions and variables that factor
> in to whether a file should be reinstalled or not, and most of them
> have been fixed, but this patch is hiding all of the rest of the bugs.
>  I'm not sure if you got the email I sent Jer about this problem, but
> the real bug is that we aren't handling the REINSTALL variable
> correctly (or REINSTALLMODE for that matter.)  Can you please describe
> the exact case for why this patch is needed?
>