PDA

View Full Version : Q104431 Update installs



NewsArchive
11-10-2000, 01:00 AM
Hi Sean thanks for replying.

I may have ended up going down the wrong path....

The first CostFinalize is executed in the UI before even the InstallWelcome
dialog is displayed.

Pretty much all of my Features/Components have no destinations at this point,
I have a CA that creates them (from user input - i.e. IP addresses, UNC paths
etc.), and I schedule this to execute just before the second CostFinalize in
the execute sequence - the one that Win98 apparently skips by default.

I read/tried to implement Q108845 as it relates to this issue (only it
concerns itself more with evaluating conditions on features/components), but
to no avail.

A bit more research has suggested a simpler workaround - manually
calling/executing the CostFinalize action, to ensure that it gets called
regardless of which OS the install program is running on.

I've also read that if you do this via a InstallScript CA, that the install
can hang, and so it is better to manually trigger the CostFinalize action via
a VBScript CA instead.

Can anybody confirm any of this?

Thanks in advance

Roy


Sean Kennedy wrote:

> Are you executing CostFinalize before you set the new directories? On 9x I
> would think this might cause it not to evaluate your new <INSTALLDIR> for
> all of the components, but it would work under NT.
>
> > I have my WI routine, which works fine on NT, but not on Win98.
> >
> > Any help, especially from InstallShield, would be appreciated.

pspsps
01-31-2002, 02:28 PM
I have been using the update technique in

Q104431, "INFO: The Creating Updates/Maintenance Packs Sample"

and I think its a very weak example. IS support keeps recommending this, and its seems too simplistic to work reliably.

Updates depend on an integer that you build into your install, and write into the registry. Then you switch on a comparison of the number when testing for an update.

Somehow I don't think a major software product distribution could rely on this.

singletd
02-01-2002, 12:53 PM
I agree with you.

I have used this method to patch an installation by adding anew component and it did work okay until the user went into modify mode from Add/Remove programs and the new component was not there to be uninstalled?

I assume that ceratin files such as Data2.cab in DISK1TARGET will need updating to handle this?

Maybe InstallShield can recommend which files should be updated and the best way of how to do this as well, so that the patched installations Modify and Repair modes work correctly from Add/Remove programs.

DS

hellokarl
11-16-2003, 10:34 PM
Q104431 is a very simplistic update mechanism. We tried it on a sample release and found the update key is not so much the problem as is missing files. This breaks the update in several ways;

When a user makes the install of the original system on a test server and then moves all the files to the production server, the update fails on the test server because it can't find any files to update.

When a user makes the installation of the original system, and then the update has more files than the previous install, the extra files are not installed.

Anyone got any references to how to improve the whole update mechanism?
(we've "upgraded" from 6.31 to 9.0 and now the project file doesn't work at all anyway - but helpdesk is working on that one).