I am having troubles uninstalling a shared win32 SxS assembly using Wix3 on WinXP. My wix file looks pretty much like the one described in http://n2.nabble.com/Tutorial-How-to-install-files-into-WinSxS-td841475.html.
The problem is, if a module from the assembly is in use, the uninstaller completes with a success errorcode but leaves the assembly in a broken state: the manifest and security catalogue have been deleted, but the assembly directory still exists with my dlls in it. The DLLs survive reboots, so are not marked for deletion in any way. There are no obvious errors in the MSI log file.
If I try re-installing the package, it skips installation of the assembly. The components are no longer registered, so I can't get Installer to try uninstalling again. The following is shown in the msi log on subsequent installs:
MSI (c) (98:44) [11:46:56:263]: skipping installation of assembly component: {26A273E7-7F9A-4F77-9FA8-5E413A155BEC} since the assembly already exists
I can't find a way of bringing the SxS back into a good state short of manually deleting my assembly's directory, which at this point is no longer being protected by XP System Restore.
While my actual installer is much more complicated, I have been able to reproduce this with a very basic installer that has a single feature containing a single component. That component contains a dll, a manifest and a security catalogue. The dlls <File> element has KeyPath, Assembly and AssemblyManifest attributes set. Once installed, I can simulate a process loading the assembly by opening the DLL in MSVS.
Is there an extra action I should be invoking on uninstall to detect and prevent this situation? Alternatively, is there a way of forcing the install operation to be performed even though the DLLs already exist in WinSXS?
Somewhat related, I also have a problem if I try to upgrade the package that installed the assembly if the assembly is not changing. The installer decides that it does not need to reinstall the assembly (since it has not changed), but then when the SxsUninstallCA runs, it goes and removes the assembly. From Installer's point of view, the component is still installed, but the files are gone.
I'm working on something which I believe I'll run into this same problem.
My assumption at this point is that I’d have to solve this with a little mojo from the HKLM\SYSTEM\CurrentControlSet\Control\Session Manager\PendingFileRenameOperations regkey, and a possibly bit of file permission smacking (if the files were still locked).
Which of course, requires a reboot.
Of course, in your installer, you could check for the presence of the keys and block on install until they're gone. (ie "you need to reboot").
Related
Good day,
I have an application that I developed that transfers files between two machines ("site" and "server"). This application was set to target dotNet 3.5. Furthermore, I am using Renci.SshNet to handle the connections between the machines and the transferring of said files.
The issue that I am facing currently though is that about 70% of the "site" machines do not have a standard dotNet and is also quite old; thus these machines do not support all the required functionality as the external dll makes calls to System.Threading.WaitHandle.WaitOne() and System.Threading.WaitHandle.WaitAny(WaitHandle[], Int32) and other overloads of these methods.
The workaround that I have for this though is to install netfx20SP2 or netfx30SP1, yet I am not in the position to perform this update on all machines as they are scattered across the country and have data limitations (bandwidth and cap).
What I want to do possibly is to embed the System.Threading dll that I have downloaded and then the application should use those classes instead, or alternatively just point the application to use the said dll.
Is this at all possible, or do you have to load the dll into the GAC? And also, will it be possible to "run" this higher version of System.Threading in the application while the system itself is on a lower framework version. Something is telling me that the best bet will be to actually run the service pack installation to avoid unnecessary coding but I'm not sure exactly how to approach this.
Thank you in advance for any assistance / suggestions,
JD
To allow the execution of an application that, let's say, targets .Net 4; while the machine itself only has let's say, .Net 3.5, installed, one can redirect Windows to check the local (executing) directory for dlls that should contain the required symbols loaded into memory instead of the default symbols that get loaded upon execution (the default would be the NetFx installed on the machine - which I believe the highest version of the framework that can be found upon loading when the execution starts or would be the highest available version that is lower or equal to the targeted framework).
This file's contents (myApp.exe.local) are ignored. It is just there to tell Windows to
look in that folder for the applicable symbols and if not found, the system will roll back to attempt to load these symbols from the NetFx directory.
Read more at Microsoft Dev Center - Docs (link is attached to the following paragraph which is a Copy-Paste of a section of this document).
To use DLL redirection, create a redirection file for your application. The redirection file must be named as follows: App_name.local. For example, if the application name is Editor.exe, the redirection file should be named Editor.exe.local. You must install the .local file in the application directory. You must also install the DLLs in the application directory.
I'm getting the following error when attempting to install a nuget package into a standard .NET framework 4.7 project:
The specified path, file name, or both are too long. The fully qualified file name must be less than 260 characters, and the directory name must be less than 248 characters.
I'm using Visual Studio 2017 15.3.3 Enterprise (latest and greatest).
Given that this is my package, I have full control over the source code. The interesting part is that I have used this package in the past with no changes in the name, but for this go round, I rebuilt it to add a feature and am now getting this error.
Even more interesting is that I have packages from the same library, with the same namespace conventions, with longer names, that work just fine and have installed into this same project with no problems at all.
I have already tried shrinking the package name, shrinking the class names within the package itself, cleaning out the build directory, cleaning out the package home from the nuget server (it's a local server with the latest nuget.server installed which otherwise works just fine), and even clearing out the bin directory of the project in question, clearing out ALL the bin directories of ALL the ancestors to the "offending" package, clearing out the package cache, rebooting the computer and rebuilding the entire nuget package chain from scratch, all to no avail. I was told by one of the MS MVP's that "they fixed that". Apparently not.
Any help would be appreciated here, I'm at my wit's end and have run out of ideas to try.
Thanks.
OK, big Thank You #danmosemsft who suggested digging around with the SysInternals process monitor. After fiddling with it for a bit, I finally figured out how to narrow the result set to just file activity. What I noticed, and the nuget engineers should TAKE NOTE of this: The problem was NOT a too-long project name, rather, nuget was attempting to update a package that was no longer there. Why it went away is a mystery yet to be solved. I normally stay out of the packages directory and do not fuss with the packages.config file. I think that this might have to do with my impatience waiting for VS to start, load all the goodies and then allow me to perform a "Manage NuGet packages" - update all. I remember seeing an update to either NUnit or FluentAssertions that wanted to perform some additional file activity aside from just installing the next version, a script I believe. Can't speak to it with assurance, I wasn't paying that much attention as third party updates usually "just work". I didn't see the "finished" line from NuGet so I think that was the root of my problem. Rather than wait until VS has settled down, I pushed it a bit (hey, the buttons responded so there shouldn't be any problems...).
As a result, the packages directory was absolutely chock full of old stuff that did NOT belong there. So, I manually cleaned up all the cruft, manually cleaned up the packages.config file, restarted VS, waited for it to settle down, performed my NuGet updates and viola! no problem - HAVING NOT CHANGED ANY OF THE ANCESTRAL PACKAGE NAMES BY EVEN A SINGLE CHARACTER.
So, what do I conclude from this? It is my belief, and the guys that actually build nuget and nuget.server should take a closer look at the errors being thrown, such that I think that the error is not so much a path too long error, rather it's a "hey, I didn't find the file I expected so the file name is full of junk (and probably too long now) so I'll throw an error that says it's too long and quit". It's seemingly a failure to handle a missing package/package directory that is causing this particular problem
I solved my problem by ensuring that all the package directories were clean of all junk and rebuilding from a clean source. My problem is now solved.
Thanks to all of you that responded.
Update: While the above contributed to the solution, it was NOT the answer. Here is the sequence of events that led to this problem and it's ultimate resolution.
The solution was created in the C:\User\Sam\Documents\Visual Studio 2017\Projects directory with the specified name of AWE.Lib.ADO.MsSqlSvr.ServerEntityHandler. This worked just fine, no errors. However due to a change in naming scheme from on high, the root directory for this project was changed from "C:\User\Sam\Documents\Visual Studio 2017\Projects" to "C:\User\Sam\Documents\Visual Studio 2017\Projects\DotNet_4.7\AWE 8.x". No problem, I thought - given that a co-worker who also happens to be a MS MVP had told me that all naming length restrictions had been removed in VS 2017. So...I moved the project from it's current home to the directory specified. Compiles just fine, brings in UPDATED BUT ALREADY INSTALLED nuget packages just fine, etc.
Or so I thought. When I needed to add a NEW (one that had not been a part of the solution before) nuget package to the mix, I received the above error. Turns out that the new name of the receiving solution is a few characters longer than VS will accept - the naming length restrictions are STILL IN PLACE.
How did I finally solve the problem: After struggling with this, I threw my hands up and decided to start all over again - a true File | New. So, I started with a new solution named as follows:
"C:\Users\Sam\Documents\Visual Studio 2017\Projects\DotNet_4.7\AWE 8.x\AWE.Lib.ADO.MsSqlSvr.HndlrServerEntity"
THIS GENERATES AN ERROR - name too long. I wondered at Nuget's error in that it specifies that the name should be less than 248 characters in length or 260 maximum.
What I am allowed to use iaw the new solution dialog is this: "C:\Users\Sam\Documents\Visual Studio 2017\Projects\DotNet_4.7\AWE 8.x\AWE.Lib.ADO.MsSqlSvr.HndlrServerEnt", for a total of 106 characters in length. If the directory is shortened, I can add to the length of the name. If I shorten the length of the actual solution name, again, VS'll accept it. So long as the total length of directory plus solution name is less than or equal to 106 characters, there isn't a problem.
The nasty bit comes from creating the solution in one location and having it work in all respects just fine, moving said solution to a different directory, still having it function in all respects (I did NOT need to add any new nuget packages yet), then trying to add a new nuget package to the mix after the move. THAT is what triggered the above nuget error.
So...the ultimate "fix", use a shorter name as it seems that 106 characters is the limit despite what the error messages are saying (and what the MS MVP was told/told me).
There is another reason for this error message by the compiler.
While building , make sure the source code is placed at a folder location which is less than 260 characters long.
For example, a path like C:\Users\User\source\Services\Exp\Sample-web-application-indot-net-displaying-RestAPI\Sample-web-application-indot-net-displaying-RestAPI\SportsStore is around 150 characters long but there are sub folders in the solution which in turn have source code files and so on.
Sometimes the overall length of the path of some files breaches the 260 character length.
I think the future versions of Visual Studio would have a bigger length allowance. Until then, we can make sure that our file names are not too long.
I was running into the same issue after I moved a project to another folder. In my case I closed VS renamed the .vs folder in root to 1.vs (effectively removing it) and re-opend my project.
In my case, I was first trying to install a package using Manage Nuget Packages for Solutions and was getting this error. Then I tried installing same package using Package Manager Console and it worked fine. I again uninstalled that package and tried installing using Manage Nuget Packages for Solutions and this time it worked fine as well.
Well in my experience all i had to do is move the entire project to my c: drive, delete unnecessary folders to ensure the path would be shorter. Done deal.
This error comes to me when I tried to copy the project folder to OneDrive and the problem is that OneDrive is not uploading long name files.
I've fixed this issue by just copying the project folder then paste it in the new laptop using USB.
I wish this could help
Move the folder of the project to a folder within a few levels from the root dir. For example the Desktop and voila.
Try to close Solution from File->Close Solution and opening it again.
For my case uninstalling, installing, or even updating NuGet packages, nothing was working but reopening(sometimes you can also close and open Visual Studio again) the Solution did the magic.
We're using TeamCity (9.0) as our CI server to build, test and deploy several applications. Recently we are seeing occassional (one in every 30/40 builds or so) NuGet (2.8.3) errors as follows:
[restore] The process cannot access the file 'C:\BuildAgent\work\e32cbd0940f38bf.....\packages\Newtonsoft.Json.5.0.6\Newtonsoft.Json.5.0.6.nupkg' because it is being used by another process.
where the actual package seems to differ from time to time.
We suspected it has something to do with the same package being referenced in multiple projects within the same solution, but I would expect NuGet to be able to handle this correctly by filtering out duplicates instead of attempting to retrieve the same package multiple times, thereby ending up with write-locks when restoring the packages to the work folder.
As a first step of each Build Configuration we have a 'NuGet Installer' step set to 'restore'. I've tried fiddling with its settings (different 'Update modes', '-NoCache', older NuGet version (2.8.0)), but to no avail.
Has anyone else experienced similar issues, and if so, has any suggestions on how to ensure this error does not occur.
Any help would be greatly appreciated!
I had the same issue with Jenkins and fixed that by adding "-DisableParallelProcessing" to the nuget restore command, the final command would look like:
nuget restore "%WORKSPACE%\Solutions\App\App.sln" -DisableParallelProcessing
Excluding NuGet package files from our anti-malware products resolved this issue for us.
I used the SysInternals Process Explorer utility on the build agents to search for file handles for any *.nupkg files while the builds were running. After several builds I observed the anti-malware products briefly locking these files during the NuGet restore operations. Adding an exclusion to the anti-malware scanning rules prevented these locks as the files were no longer being scanned.
In our environment we use two different anti-malware products on different build agent servers. We encountered this issue with both products.
As far as the error message is concerned, I also came across it.
I debugged the “nuget restore” process, breaking at the point where the .nupkg is copied to the local repository, and then freezing the thread while the file was opened for writing. And sure enough I got the exception in another task, due to the fact that the two packages had Ids where one was a prefix of the other. I filed an issue for this : https://nuget.codeplex.com/workitem/4465.
However, this is probably not exactly your problem, since the error in my case is on reading the .nupkg of the package with the “long” name, and I don’t think there is a package with an Id that is a prefix of NewtonSoft.Json (whereas it is very possible the other way around : there are for instance NewtonSoft.JsonResult of NewtonSoft.Json.Glimpse).
I installed new Newtonsoft.Json and problem disappear
You can turn on build feature Swabra with option "Locking processes" (requires handle.exe). And check are there any files locked after build's finish or not.
If there are no locked files then try to run Nuget using command line build step instead of NuGet Installer. If the issue is reproduced then most probably it means that the issue is related NuGet.
I have an MSI package I maintain with Installshield 2012 Pro.
The package was created a few years again and requires periodic maintenance as new versions of the application is installs are released by our development team. This MSI has been used in many locations, on many machines without issue.
At one customer location, they've install about 20 copies of the application on 20 machines. Each machine is built exactly the same.
Periodically, for an unknown reason registry entries relating to that application go missing. Not all of them, a select one or two... the same entries always go missing.
We so far have failed to track the exact root cause, however one of our support engineers noticed the MSI repair dialogue appearing when they opened the application shortcut. Following this, the registry entries went missing.
On each instance of the registry entries going missing we have removed the MSI from the c:\Windows\Installer directory. On machines we've made this change too, the registry changes so far have not disappeared again.
This leads me to think the MSI is somehow removing the selected registry entries during the repair, but I don't understand how.
There are at least 50 registry entries under the same key as the entries that get removed. (these are created by the application not the installer)
The MSI package doesn't insert any keys at all during install
There are no custom actions at all
I only have one key file set in the whole MSI, and this is for a component installed in Windows\system32
None of the shortcuts are 'Advertised'
There appears to be a clear link between removing the MSI from the Installer directory and no further occurrences, but I fail to see what in my MSI could cause specific registry keys to disappear.
I realize the exact reason this this is subjective to my MSI or the customers systems, so as generally as possible I ask:
Why would Windows launch the MSI repair function, considering the above
Can an MSI alter the registry, considering the above
Is there anyway post installation of an MSI to disable the auto repair function for this MSI?
How should I alter the MSI to ensure auto repair doesn't occur?
Thanks
Start by proving whether the auto repair is related to your product. Check the application event logs for the component id that caused Windows Installer to kick off the auto repair. Then, if it's part of your installer, figure out why.
It seems unlikely that your MSI would alter the registry. Although you do not specifically say there are no RemoveRegistry table entries, it's hard to imagine any snuck in accidentally.
There are only bad ways to disable auto repair (some global, some involving not registering your installer). Avoid them.
First figure out what's wrong. Then figure out how to fix it. For instance if in step 1 you determine the component triggering auto repair is part of your install, perhaps set the logging policy on a machine where this occurs so you can get a verbose MSI log.
I have a setup project with InstallShield 2010. I'm deploying a configuration file during installation. However, when uninstalled, InstallShield decides to delete it (which is normal).
The question is, is there a way to keep the file on the hard disk even after the application in uninstalled? I don't want to reconfigure the application every time the user uninstalls/installs.
Edit: I'm using MSI project.
You don't say what project type you are using. Either way, put the config file in it's own component. For MSI projects, set the Permanent attribute to true. For InstallScript projects, set the Uninstall attribute to false.