I am working on a VB6 application which has many executable and an Active X dlls.
And there are to be updated in c;lient machines to lates version once in a while which i am asking the user to update manually.
Can you please suggest me a way using which i can update it automatically from the files that can be available online.
Thanks.
Windows Installer has features supporting Patching and Upgrades. Using those techniques you can create various levels of "upgrade" packages.
Your application would need a separate "update" utility that is spawned when the user approves updating, perhaps in response to a prompt your program raises after checking for new versions.
This updater would check the current version and the remote site's catalog of updates to pick the appropriate package, download it to a temporary location, start Windows Installer to process the package (or packages, sometimes you might need to run several Installer runs), and clean up the temp location. Then you might offer to restrt the updated application or on some occasions need to reboot.
This updater would be a fancy form of the common "installation bootstrapper." As you can tell it needs some "smarts" in order to tell what package or packages to download and install in what sequence, when it needs to request rebooting, etc. This would probably be based on a downloaded "rules script" it obtains as part of selecting a valid update option.
After all, sometimes you can just apply a minor upgrade or patch upgrade, sometimes you need a more complete install or entire reinstall.
If your needs are extremely simple (just an EXE and maybe a few DLLs and OCXs - preferably using reg-free COM) you may not need to go to these lengths. However when you start adding in other considerations like multiple programs, data directory creation and security settings, possibly running a settings file conversion or even database conversion, DCOM, firewall, etc. configuration, database drivers or providers, etc. things get complicated quickly. Too complicated for simple snatch and grab updating.
And admin rights/UAC issues are a factor so you'll probably have to deal with privilege elevation.
None of this is trivial stuff. There are people who do little more than construct and test such deployment systems as their entire job.
If you use soemthing like Inno setup to install the application then an update is simple a matter of running that periodically.
You can either detect there is a new version available by checking a web site/local server, or just prompt to run the update after X days.
Related
Thank you in advance for considering this question. If a similar question existed, I was unable to find it.
The Issue: Our company packages an application into an MSI. This MSI when installed outside of any GPO properly updates, blocks attempts to downgrade (or move from a higher revision to a lower revision), and never has trouble uninstalling previous versions of the application regardless of how long ago those versions were created/installed. For example, we can install version 1.2.3 and then install version 2.3.4 and the application will properly install without issue. However, we work with a customer who uses GPO to deploy our application to hundreds of PC's. Each time we have provided an updated version of the application the following has been indicated:
On any machine where a previous version of our application was installed via GPO, no matter what the previous version is, the update successfully installs without issue.
On a machine where the application was manually installed (outside of the GPO), and an attempt to update the application via GPO is made - either the application is installed in addition to the old version, OR there remain registry keys to the previous version of the application and the application cannot open/run correctly. In this case the registry keys must be manually removed, and the install is then attempted again from a clean machine.
What we know is that on any machine where the application was originally installed via GPO - updating the application is no problem. On every machine where the application was not installed with the GPO in the first place, updating via GPO fails with one of the problems presented above.
My question is: Is there a technical issue with how the installation is being handled partially through the GPO and partially outside? Does the GPO need to be responsible for the entire life-cycle of the application? OR is it a reasonable expectation that the application be updated both on machines where the original version was manually (outside the GPO ) installed, and when it was installed initially from within the GPO?
One solution we are aware of is simply having all computers manage the application life-cycle (since we know updates work in that environment already), however this would mean that many computers would need to have the manually installed versions removed by hand - and then properly handle the installation through GPO which is an extensive bit of work.
We would greatly welcome any solutions, references to technical documentation that formally shed light on the proper management or expectations here, or links to information. Our research suggests that it is "best" to manage the entire applications life-cycle inside the GPO - but I have as of yet been unable to determine that it is 100% necessary to do so.
Looking forward to any assistance. If any further technical details are required to help the viability of the question, please don't hesitate to request such details.
If you end up with two versions installed in Control Panel, then all other things being correct, the most likely explanation is that you upgraded a per user install with a per machine install (or vice versa). In the GPO world that's related to assigning it to a user or to the computer, something like that. That's easy to verify by getting a verbose log and checking the FindRelatedProducts actions for an indication that another product was found but in a different context.
When you're in GPO mode all the time, I assume each one (whether it's per user or per machine) is consistent, therefore upgrades always work, but they don't work cross-context.
I believe GPO suppresses the UI in most cases, and the UI (or the UI sequence) is sometimes where per user/per machine is set. That might be something else that would cause it, depending on how the GPO publishes to the computer or the user.
I created a Visual Studio 2010 installer project and added the postgresql-9.2.401-windows.exe file to it. In the "Commit" custom action, I added that file, removed any arguments and set InstallerClass to False.
Now, when I run my application's installer, the PostgreSQL setup starts up, shows its splash screen and then jumps to 100% CPU usage and starts leaking memory for a long time (certainly longer than a few minutes), after which it bails out. Running the .exe on that computer works just fine.
How should I package PostgreSQL in my installer?
Personally I don't like programs that silently install the standard PostgreSQL for Windows. It's deeply annoying to get mailing list posts from people who're asking "What's this PostgreSQL thing and why is it using 88 bajillion megabytes of RAM, I didn't ask for it, remove it now!".
(Yes, I'm mostly talking about a certain annoying gambling card-counting program with the initials PT whose name I will not mention directly).
If you're going to bundle PostgreSQL it'd be great if you could use a non-default port and non-default data directory. If you install a service to auto-start it, don't use the default service name postgresql-[version] but something like myapp-database or myapp-postgres.
This is all done much more easily if you just bundle the PostgreSQL binaries in your installer then create the service yourself with your installer tools or via sc.exe. You may choose to simply pg_ctl start/stop the server from within your program, though this requires some thought about access permissions to the data directory.
The downside is that when there are PostgreSQL updates you need to update your installer. On the other hand, you probably don't want people upgrading PostgreSQL independently of your software updates.
I suggest using the .zip binaries from here and bundling them into your installer directly.
If you really must use the executable installer the documentation contains instructions on silent installation.
I'm trying to release a program with a client/database approach.
The client and database can be installed on 2 different computers.
What I'm trying to do is an installer that at the beginning asks if you want to install the client, the database or both.
Is this possible? Which section should I modify?
I warn you, I've never wrote an Inno Setup script, I always used ISTools, but if there is no other way, I'm ready to get my hands dirty :P
Edit: a bit more info.
At the moment I have 3 installers, one for the client, one for the DB (which is postgreSQL) and one for PDFCreator.
What I'm doing is copy the files to a directory and then ask at the user what he wants to install: [Run] section, Post Installation.
Yes, it is possible. Take a look at the CodeDlg.iss example. It show how to create a dialog to ask which edition to install. In the case of CodeDlg, it is various editions of a program (light, sponsored, paid) but the concept is the same. I've done this before as well. In my case, during initialization, I read a registry setting and set some variables up to choose which version (client or server) I install. I don't have it here but will get it later and add to this answer.
I know on windows there are a bunch of installer tools you can use to create an installer, but on Mac OS I've seen two ways to install apps:
A DMG file which you download, double-click, then run an application inside - the application typically has you drag an icon to another icon (representing the Applications folder) to install the app
Another type of file which launches an apparently standard installer, which sometimes brings up a warning like "This installer may run a program to determine if you can go ahead with the install"
What's the "standard" way of packaging an app for install on Mac OS? Is one of the above the Apple-recommended way?
Thanks.
Apple is very clearly making the "standard" to be downloading a program from the App Store. This has the benefit of making application installation transparent to the normal user. And, believe it or not, normal people have a lot of trouble with the concept of installing a program. Of course that benefit comes with some costs, but this isn't the place for that debate–there are plenty of other places for that.
Assuming you don't want to or can't go the App Store route, both PKG and DMG are common ways to distribute a program. Use a PKG if you need to install files aside from your application bundle (which should not be a common use case). In all other cases use a DMG that prompts the user to copy the application into the Applications folder. But a lot of your users will not understand that they need to do that (unless your target audience is solely knowledgeable computer users). They will run your application from the disk image. Ideally in this case, your program will detect that it is running from a disk image and offer to copy itself into the Applications folder.
Packages works well. If your deployment process must be kept simple, it is great.
The Quick build consists of dragging your .app onto Package and it is done.
For advanced packaging, you can also provide a certificate.
http://s.sudre.free.fr/Software/Packages/about.html
We're discussing two things:
first and most importantly, the standard method by which the bits of an executable get laid on the disk in a way that's accessible and properly registered by the system
second, the mechanism for preparing the .app, as recommended by Apple
An older marketing page on Apple's site says it's recommended to create packages (so the Installer application can move the bits in place) with the PackageMaker application.
Its usage is described here: mactech.com/articles/mactech/Vol.25/25.03/2503MacEnterprise-PackagingforSystemAdministrators/index.html.
But as others have mentioned, the elephant in the room is the MacAppStore(MAS for short). Until it's debut, what was standard for large companies was their own custom scripts rolled into an older-style 'bundle' package or using an executable like the VISE installer. Smaller developers usually tried to make their app installable via drag-drop, distributed in zip archives or disk images(for simplicity's sake).
The MAS is different: as of 10.7 it uses a package format (which debuted in 10.5) referred to as a flat package (really a xar archive, explanation here)
which is transferred over http to a hidden folder, installs directly to Applications(after which the temporary folder it is downloaded to is deleted). It drops its receipt and a bill or materials file into /private/var/db, and is therefore audit-able by the built-in command line pkgutil tool, described here: mactech.com/articles/mactech/Vol.25/25.12/2512MacEnterprise-PackagesReceiptsandSnow/index.html
A benefit of using the flat package format is you can pull things over the network more safely and efficiently, but it isn't as easy to work with as bundle packages if you are testing and modifying the package regularly, or iterating to ensure scripts that perform actions or checks work well. Even when flat, putting the pkg in a archive or disk image is recommended for flexibility. More distribution tools expect DMG's than zip's, so there's that as well.
Besides what Apple recommends and what is standard, common practice, there's this article: https://www.afp548.com/2010/06/03/the-commandments-of-packaging-in-os-x/ which discusses the why's and hows (although mainly for system administrators) of packaging for wider distribution. It is greatly recommended to get more of a feel of how and why things go wrong, and what to avoid.
Try Iceberg!
Another one package creator.
In OS X, many applications are just created as a Relocatable
application bundles that the user just need to copy to the
/Application folder (or any other location). In other cases, when you
need to perform some operations over the machine (such as adding users
or changing permissions) you can use a PKG installer (for example built using PackageMaker), which allows
executing some pre and post install scripts and support some basic
installation configuration, like selecting the installation drive.
Sometimes, as with complex server software, you need more flexibility, for example to show custom pages to
the end user requesting information required to install your
application, like the MySQL port and password or proxy information to
download requirements on the fly (or simply to make it look fancier
:)). For this cases there are other installer solutions like our BitRock
InstallBuilder (disclaimer, I'm one of the developers).
InstallBuilder also has the advantage of generating multiplatform
installers using the same project with very little customization per
platform.
As most of you probably noticed, when uninstalling an MSI package Windows will ask for the original .msi file. Why is that?
I can only see disadvantages to that:
not resilient to network changes.
not resilient to local disk changes.
unexpected by users.
typically requires users to leave their desk and start a crusade to get the correct CD.
kind of proves installations are not self-contained.
promotes the use of unsafe tools such as msizap.
which in turn promotes the "next time I'll just use a zip file" mentality.
Could someone shed some light on this?
Fix Broken Uninstall: You can try the newest FixIt Uninstall tool from Microsoft if you have problems uninstalling an MSI. And one more link: Uninstalling an MSI file from the command line without using msiexec (a plethora of different ways to uninstall an MSI).
UPDATE:
This new support tool (this tool is now also deprecated) can be tried on recent Windows versions if you have defunct MSI packages needing uninstall (rather than the outdated, deprecated, unsupported msizap.exe).
Some have suggested to use the tool linked to here by saschabeaumont: Uninstall without an MSI file. If you try it and it works, please be sure to let us know. Feedback in that answer indicates that it works (I don't have any stuck setups to test with as of now).
Why are you asked for the original installation media?:
The original MSI is not needed for uninstall unless the MSI itself is badly designed - or the cached MSI is missing (see details below).
All installed MSI files are cached in %SystemRoot%\Installer\*.* using a random hex name.
The cached MSI file is used for any maintenance, repair and uninstall operations - and it is sufficient for uninstall in the vast majority of cases.
In some cases this cached file can be missing, and then uninstall is not possible at all in some cases (some theories as to why this can happen - MSI design errors, anti-virus quarantining, system restore, tinkering, developer system in erroneous state from development work, etc...). See more info below - and links to force uninstall or unregistration of the product.
The original source is only needed if files need to be copied to disk (for a maintenance install), or the MSI does an explicit request to resolve the original source via the standard action ResolveSource or via a custom action (which shouldn't be done in a properly authored package - I think the MS Office package contained this ResolveSource error back in the day, causing everyone to go looking for their installation CDs/DVDs).
In previous editions of Windows this cached MSI was stripped of all cabs, and hence contained the installer structure only, and no files.
Starting with Windows 7 (MSI version 5) the MSI files are now cached full size to avoid breaking the file signature which affects the UAC prompt on setup launch (a known Vista problem). This may cause a tremendous increase in disk space consumption (several gigabytes for some systems). Check this article and especially the discussion at the bottom for more intel.
To prevent caching a huge MSI file, you can run an admin-install of the package before installing. This is how a company with proper deployment in a managed network would do things, and it will strip out the cab files and make a network install point with a small MSI file and files besides it. Note that this may yield a UAC prompt is some cases since the extracted MSI file is no longer signed - this must be tested with your SOE / desktop configuration.
Read my answer in this thread for the full description of admin installs: What is the purpose of administrative installation initiated using msiexec /a? or this similar but perhaps more accessible answer: admin install and its uses
In some rare cases the cached MSI (with the random name) can be erroneously missing, and uninstall will then ask for the original MSI in order to complete the uninstall. This does not happen often. It used to be the case that one could use MsiZap.exe to clean out such an install, but this tool is outdated, deprecated and unsupported. Don't use it - there are too many incompatibilities with newer Windows versions and you create new problems. Perhaps try this support tool instead (also deprecated). The only option I can suggest now is this answer from saschabeaumont. If you try this tool, please let us know if it works for you. If you want to figure out what could have caused the cached MSI to be missing, try to read section 12 here: Uninstalling an MSI file from the command line without using msiexec (in short potential causes range from interference with system restore, anti virus and cleanup scripts, to erronous manual tweaking, low disk space, power outages, developer box debugging errors, badly designed MSI files with duplicate package codes, failed patches, etc... Many theories, few certainties I am afraid).
As a last resort you can try system restore (unless it has been disabled entirely or partly) to go back to a previous installation state and see if this solves your uninstall problem (you can find video demos of this on youtube or a similar site).
Be aware that system restore might affect Windows Update that must then be re-applied - as well as many other system settings. I have seen new, unsolvable installation problems resulting from a system restore, but normally it works OK. Obviously don't use the feature for fun, it's a last resort and is best used for rollback of new drivers or setups that have just been installed and are found to cause immediate problems and such issues. The longer you go back the more rework you will create for yourself. A lot of self-evident stuff, but I guess it needs to be mentioned.
Since I mentioned system restore I suppose I should mention the Last Known Good Configuration feature. This feature has nothing to do with uninstall or system restore, but is the last boot configuration that worked or resulted in a running system. It can be used to get your system running again if it bluescreens or halts during booting. This often happens after driver installs. This will do nothing to fix your failing uninstall though (or I would be very surprised).
Related answer:
Uninstall without an MSI file
In addition to this answer, perhaps this article on various ways of uninstalling MSI packages is of interest. It is a rather popular article with a high number of views:
Uninstalling an MSI file from the command line without using msiexec.
There are a few reasons for keeping the original msi:
The uninstaller uses it to know what files and registry keys were installed and make sure they are all cleaned up.
The msi may contain special uninstall actions that need to be performed.
It allows you do to a 'repair' operation from the Add/Remove Programs menu, regardless of whether or not you saved the install file yourself.
The normal way of things is for Windows to keep the file cached for you, so you don't have to think about it. See your %WINDIR%\Installer\ folder. The only reason it would ask your for the original msi is if something is wrong with the saved file. This addresses most of your concerns, though it does raise a new one (disk-space).