To my astonishment I found VB6 code that uses Declare statements to define functions in a .dll that lives in the Program Folder without it being registered on Windows. This seems like a supersimple way to avoid the .dll hell without having to resort to using Side by side manifests. Can I read some more about this somewhere? Are there snags?
The Declare statement is used to do "just in time" binding to non-ActiveX DLLs. Until your program "touches" a Declared entrypoint no attempt is made to load the library.
It basically has nothing at all to do with the topic of DLL Hell.
Muddled thinking can even lead people to plop ActiveX DLLs "next to" the EXE which actually can result in DLL Hell because people who tend to do this also use poor techniques for installing and uninstalling applications.
Poorly designed application A deployment plops a commonly shared DLL or OCX next to the EXE.
Poorly designed application is run, the VB6 runtime can't find the classes in the registry, does a DLL Search using Windows heuristics, immediately locates the DLL next to the EXE and calls its self-registration entrypoint.
Innocent, properly designed applications B, C, D are later installed that use the same DLL/OCX and their installers find the library already registered.
Poorly designed application A is uninstalled, typically by simply deleteing its folder in Program Files.
Applications B, C, and D (and any future applications using the library) are now broken - due to orphaned component registration pointing to a non-existant library.
Moral of the story:
Never, never, never put DLLs "next to" your VB6 application on installation. If you have private DLLs that are not shared with other applications even then put them into a libs, etc. folder under the application folder. Possible exception might be non-COM DLLs, such as those you expect to use Declare with.
There is also a great deal of misunderstanding about manifests, of which there are multiple kinds. The ones you are probably thinking of are application and assembly manifests.
These can be used for selecting among different versions of a library installed Side by Side, or they can be used to isolate applications and assemblies which is the part that has bearing on DLL Hell.
Of course application manifests can be used to specify quite a few other things about how Windows should run the application.
Windows searches in a well-documented sequence of folders for LoadLibrary (which VB6 uses behind the scenes to resolve Declare declarations). Since the first location on the list of search folders is the app's own folder, your discovery makes perfect sense.
It doesn't resolve the "DLL hell" issue for the most part, though. It can't work for system DLLs, for instance, because Windows preloads most of them. Also, if a DLL is already loaded into memory, Windows may use that copy of the DLL (not sharing data, but code can be reused).
That's part of the reason that manifests were created; they allow an application to strictly define required versions of system DLLs in order to provide certain functionality. VB6's technique is old fashioned (just like VB6).
Related
We have a project here written in VB6 (yes, I know...) which consumes some ActiveX objects also written in VB6 (OCX files).
Recently we started using a build server for this project and, with the help of the MSBuild Extension Pack we devised a build process which compiles the libraries, controls and executable all in the correct order, registers them as appropriate and then unregisters everything to leave a clean machine for the next run.
The problem is that each compilation run leaves entries in the registry for the extended type library/object cache files which the VB6 compiler appears to require, and these entries are not removed when the controls are unregistered.
i.e. for every registered component which has a CLSID entry pointing to the OCX, there exists, after we compile the project with VB6, a newly generated CLSID for the OCA.
Because we unregister the OCX controls after each build and the OCA CLSIDs are generated afresh each time, the number of OCA entries continues to grow.
Does anyone know what causes these entries, whether we can remove them through some kind of unregistration process or whether we can prevent them being created in the first place?
A few notes :
The build machine is Windows 7
Binary compatibility is enabled, so the CLSIDs of the OCX controls aren't changing.
Reg-Free COM isn't currently an option
We cannot simply register the controls and leave them as we want to run more than one build of this project.
The key is to stop trying to treat DLLs (including OCXs) like statically linked libraries. Your build process is inappropriate.
While VB6 supported a crude "project group" capability, this was only meant for small scale unstructured development efforts like those used by under-the-radar business unit coders.
Instead strive to keep application-centric logic within your EXE projects. Factor out longer-term reusable logic into DLL and OCX projects that you treat as separate, unique internal products. Maintain these (including source control) compeletely separately from consuming applications.
Normally these should be far more stable than application code where changes are driven more frequently by the business logic they contain.
Even when some of these must contain business logic, you still want to treat them as separate software entities. Try to keep them distinct from libraries that have more stability.
This requires more discipline than monolithic "tower of Bable" development but there are many rewards. Build time is quicker for one thing. And as a bonus almost all of these source control related issues disappear for the applications which have a much higher rate of tinkering and fiddling.
I'm maintaining a VB6 application with many COM components (DLLs and OCXs). In order to streamline development and deployment I'd like to use reg-free com. The problem with development is that the application runs within the VB6.EXE instance. How can I trick VB6 to use my (unregistered) components? It is very important for me to not have to go through registering/unregistering components when switching between branches. Generating a .manifest file for VB6 is not out of the question but is there some other, more optimal way, to specify a .manifest file when launching VB6.EXE?
Note: The Activation Context API doesn't seem to help, even if used from within the development environment.
Solutions I've thought:
A utility application that activates a context from a manifest and launches VB6 as a child process (doesn't work; processes don't inherit activation context)
Injecting context activation into the VB6 process at startup (too complicated; must hack the executable to do this)
Hosting VB6 in my own process after activating the right context (can't even find out if this is possible)
Using a VB6 Add-In or other utility that runs within VB6 to activate a context (tried that but it doesn't seem to work)
Update Jan. 16
As suggested by wqw, I did some testing with a VB.exe.manifest. The VB6.exe.manifest worked, with some caveats:
The SxS dll specified in the manifest would not appear in the references window on projects that didn't actually reference the component
On projects that did reference the component it would be shown to reside in the directory according to the following order:
The pathname recorded in the project file (if the file was still present)
A pathname as if it resided in the same folder as the project (vbp)
If the file was not in any of these folders, the project would not compile (just running the code causes an internal compile in VB6) with the message "Can't find project or library".
Obviously, VB6 actualy scans the registry to find COM components and verifies, during compilation, that they exist where they say they exist. I'm not sure what that might mean if I actually want to use VB6.exe.manifest to redirect COM component instantiation. Perhaps having dummy component files at some predefined location might trick VB6 into believing that everything is as it should be, although an entirely different set of components got loaded for use.
Further update:
I did a test on that last assumption and it proved to be false. The component has to actually be there in order for the project to compile. It must even properly load (no dummy, zero-length files accepted!). Now I'm not even sure if the manifest works. That's a more time-consuming test (requires a component with two versions that produce different results, one with the project, and one for the manifest).
Our approach to this problem was to write a build assist program that registered and unregistered components, run the VB6 compiler, and would even rewrite project files with updated GUIDs when interfaces changed. You would hand it a VBG project group and it would do the rest.
I suppose we could also have added a mode that unregistered components when you switched branches.
Are you following the practice of using "compatibility" binaries? You shouldn't use the binary at your build location for compatibility references - you should commit a separate copy to version control and configure your project to consider that the "compatible" version - only change this file when you break interfaces.
I need to build a DLL capable to load other DLLs at runtime; these other DLLs have a rather intimate relationship with the main DLL (this is Python and extensions), so they must have a common runtime. The main DLL must be a single file that can be simply copied on the target machine. The auxiliary DLLs will be placed in a different directory.
So: "common runtime" means no static linking; "single file + simple copy" rules out the shared MS redistributables, especially when coupled with "different directories".
I only see the following options: link all DLLs against msvcrt.dll; embed current msvcrtXX into the main DLL and re-export all its symbols; use the msvcrtXX of the host application. To me the first looks the simplest, because it's a common need and there are many web pages explaining how to go about it. How would you approach this?
I suggest you re-architecture so that your plug-in DLL (the one the host application loads explicitly) contains nothing but proxy functions plus the file management logic. This DLL could use a static runtime, or no runtime at all.
Put the actual functionality (and in particular all code that needs to share a runtime with the Python extensions) in a separate ("primary") DLL and put this DLL, as well as the runtime of your choice, in the same directory as the Python extensions.
The primary DLL and the runtime could be zipped to the plug-in DLL, along with the core extensions. (I presume this is within the scope of the runtime's redistributable license, since it's basically the same as the way most installers work, but you should check for yourself.)
The calls to the plug-in will be slightly less efficient since they have to go through the proxy DLL, but the performance difference probably won't be measurable. :-)
Another option (in my humble opinion, much better) would be to dynamically link - aka. LoadLibrary/GetProcAddress) the dependencies when the main starts. This would allow you to support cases of missing DLLs and/or support your own internal version scheme.
I've got a project that depends on a particular version of MSVCR80.dll (the MS Visual C Runtime) and I'm running into problems where, depending on the particular system configuration, my app doesn't always get the right version of that file. It's been a bit of a crap shoot as to what path it takes to find a file with that name, and it's not always right...
Is there a way, when creating a Deployment Project in VS2005, to ensure that my app will always use the runtime that I provided?? When I add the runtime file to the project, it asks about creating a merge module...but not really sure what that does. And regardless of creating one, the issue remains.
Martin Richter wrote an article about that on CodeProject:
Create projects easily with private MFC, ATL and CRT assemblies
This solution does not rely on your MSI packages but on the application that uses the CRT files.
I am not sure if it is your application after installation that doesn't work, or if it is a dll you use as part of the installation that doesn't work?
To make a very long story, very short: new versions of the C / C++ runtimes are installed as Win32 assemblies, or side-by-side installation. This means the files will go into folders under C:\Windows\winsxs - the Win32 equivalent of the GAC, and several versions of the same file can co-exist here.
Applications compiled with Visual Studio 2005 / 2008 will put a manifest file into the binary, and this manifest specifies what side-by-side runtime version to bind to. It doesn't matter if you put the MSVCR80.dll next to your EXE or even in system32 - the manifest embedded in the EXE will load the file from C:\Windows\winsxs.
This is all "full circle". In the old days runtimes went to System32. This caused the original dll-hell: applications overwriting each other's global runtime files. To remedy all this the idea was to "isolate changes" to each application. Hence the new approach was to isolate a local copy of the runtime file next to the EXE. Now this caused an entirely new problem: how do you make sure security updates for the isolated dll was deployed? In most cases this never happened, and you had lots of applications running with local, unsafe dll's. So what to do? The decision was to introduce the second coming of dll-hell: the side-by-side assembly approach. In this approach runtimes are not local, but global - with the critical difference of supporting side-by-side installations. This way, in theory, applications can function without overwriting each other's runtime dlls.
So that was the quick summary of "how to make runtime deployment complicated". I am not positive it is still possible to do, but did you check whether you can statically link to the runtime? Sometimes old-school really is easier...
We have started using Final Builder to create builds for our vb6 and .net projects. We are also using Visual Source Safe to manage our source. Some of our vb6 exe's are dependent on certain ocx's, such that a particular vb6 exe may require a particular version of an ocx.
The question is, should the final builder script for our exe project also re-build the ocx project, or is it better to simply pull a particular version of the already compiled ocx. My concern is that other developers could have broken the build (or created a bug) for the ocx which could then break the exe we are trying to build. Moreover, re-building the ocx project would result in the same version of the ocx but with a different date, resulting in confusion if dllhell(ocx hell) issues arise.
There is no difference in terms of building and maintaining your app between a ocx and a activex dll. The ocx should use binary compatibility and be part of your compile process.
This is however a general rule. You may have some components that rarely change if ever. In my own VB6 application I have a handful of components that reside at the bottomost level of my reference hierarchy that rarely get updated. They maybe get updated one or twice a year at best. Some haven't been updated for several years now.
However based on your description it sounds like the controls are still being modified. So I doubt the second case applies.
In the end use your best judgment.
There are two ways to use OCX/DLLs: code reusability vs. fragmentation of an over-large project.
Those meant for re-use would be absurd to build, build, and rebuild, and almost never should be customized to fit a new application. These are your crown jewels, and most people should have no ability to modify the source. They are the domain of your organization's "library writers" because that's what they are: libraries.
If you simply have large, monolithic, unweildy applications you may have to go the other route. Then OCXs and DLLs simply become an awkward extension of the "module" concept. This is why we have Project Groups.
Your library users should not be fiddling with libraries though. I'm sure they all fancy themselves able to "ensure they are up to date and performant" but that's a different debate entirely.