|
Hi,
If you use the VS wizard to create your own ATL-based COM object, the post-build step will register the object. The problem is, that registration step fails if you're not administrator on your machine (since it requires full access to the registry).
My normal day-to-day account is non-admin for security reasons. Has anyone found a good solution to this? I don't mind registering the object on occasion (with the admin account), but 99% of the time I'm changing functionality and not changing anything about the registration itself. Maybe there's a way to skip the registration process? I couldn't see a way since VS seems to do it automatically rather than with an explicit build step.
Thanks for any ideas,
DB
|
|
|
|
|
dburns wrote: Has anyone found a good solution to this?
The activity of developing software typically requires administrator privileges, get them.
led mike
|
|
|
|
|
led mike wrote: The activity of developing software typically requires administrator privileges
I'm not sure I'd agree with that statement. With the exception of registering COM objects, I see no reason why admin privileges are needed for software developement. Basically developing non-COM code requires the ability to edit/create files.
BTW I'm talking about my own machine here, so I can easily give my account admin privileges -- but I'm trying to keep my normal account non-admin so I minimize damage from either myself or worms etc.
Anyway, just looking for any tips in case someone has gone this route before.
|
|
|
|
|
dburns wrote: I'm not sure I'd agree with that statement. With the exception of registering COM objects, I see no reason why admin privileges are needed for software developement.
Well as long as you see no reason then they must not exist... problem solved.  Hopefully I can remember to avoid answering your posts in the future.
led mike
|
|
|
|
|
Hey Mike,
Sorry if that came out wrong. I was trying to be careful to phrase things in a non-offensive way, but apparently I suck at that. By "I see no reason" I wasn't saying there weren't any, I was just saying that I'm not aware of them. I intended that to be an opening for you or anyone else to provide an example. I guess I should have said "what are the reasons" instead.
led mike wrote: Hopefully I can remember to avoid answering your posts in the future
Please don't do that! It's just a misunderstanding. I do appreciate input of any sort.
Sorry for any bruised feelings.
DB
|
|
|
|
|
dburns wrote: Please don't do that! It's just a misunderstanding. I do appreciate input of any sort.
It is certainly dependent on your environment. Mine have mostly been extremely volatile and with todays speed pressures the last thing I need to deal with is privilege issues. When I want to install trial software package or start a new database instance or write and install a new Service or a GAC component or "whatever" for research I can't afford to lose time dealing with privilege issues because the paper pushing crackhead executives are breathing fire about their latest date proclamations. But that's just me. 
led mike
|
|
|
|
|
Personally, I disagree with you - when developing (i.e. designing, coding, testing software), you should have the same privileges as your target customers, which these days is more likely to be non-admin. Otherwise, you're more likely to develop apps that assume admin privs, when your customers don't have them.
I've been developing as a non-admin (with a separate admin account for when admin access is really needed) since I started using Windows 2000 in mid-2001. No problem. Seethis page[^] for useful hints and tips re: developing as a non-admin.
|
|
|
|
|
Stuart Dootson wrote: when developing (i.e. designing, coding, testing software), you should have the same privileges as your target customers
Why, are your target customers developing software? That is just retarded and not the first time I have heard it. You need machines configured as customer machines to "test" your software and installations. That is not at all the same as a developers workstation because... well it just isn't kind of like how apples are not the same as oranges even though they are both fruit.
led mike
|
|
|
|
|
Remove the registration build-step from the project and do it manually?
Alternatively, change the RGS file to register the classes in HKCU/Software/Classes explicitly, rather than through HKCR. HKCR is just an amalgam of HKCU/Software/Classes and HKLM/Software/Classes, so that should work OK.
These days, self-registering COM objects are passe - it's generally seen as more acceptable to have an installer (i.e. an MSI file install them).
PS - I agree with you about admin access. I develop as a limited user. It means I have no more privs than a user, so won't put operations requiring admin access into my apps. Big win!
|
|
|
|
|
PS - should have known, there's a CodeProject article addressing just this issue - see this[^] - just replace RegSvr32 in the registration build step with the RegSvrEx that this article talks about.
|
|
|
|
|
Stuart,
Thanks a lot for the links. I searched CodeProject before posting but apparently I suck at that too 
BTW about having the same admin privileges as your users, I tend to agree with you on the grounds that it makes for better software -- because the developers feel the pain that the users do. It's similar to the issue that developers tend to have high-end liquid-cooled hotrod systems that are 10x as fast as their customers'. Then we wonder why customers say the product is slow. Not gonna give up my machine though
|
|
|
|
|
Ah, finally got it set up the way I wanted. I had already hunted in the project for regsvr32, or any sort of post-build step, but the registration was happening intrinsically. I finally found an option "Register Output" in the link options. Set that to false and got halfway there.
On the occasions where I need to register it, I could log in as admin, but instead I just run a batch file that contains this:
runas /noprofile /user:admin_user "\windows\system32\regsvr32.exe full_path_to_control.dll"
where admin_user is the name of the admin user. I'm prompted for the password.
So now I'm happy. Thanks for the links, that was enough to get me thinking along the right lines.
|
|
|
|
|
Hey, guys.
I'm instantiating ActiveX controls in an ATL toolbar for IE and having a problem with Invoke. If I put my class that wraps the control & container in an exe and run it, then when the test ActiveX object I create fires a mouse clicked event, I get it in Invoke without a problem. The test object also shows continual x,y position in its display window, so I can see it's getting mouse events.
However, when I use this same class in the dll which loads as the IE toolbar (IDeskBand stuff, if you're interested), creating the same ActiveX test object, Invoke isn't called when I click. The control behaves properly otherwise (I don't believe it's IE security stuff or the control wouldn't load and run). Still see the x,y updates from mousemove in the object's display area, and objects I create are responding and taking the appropriate internal action when I click the mouse, but no events get fired to Invoke. Curiously, however, I can explicitly call Invoke on that object from within the code and it works just fine, passing the parameters, etc.
So, I'm doing the standard QueryInterface for the connection point container, calling FindConnectionPoint, and then making a call to Advise(). Invoke gets called for the click event when compiled in an exe, but not in an IE toolbar dll. Anyone have an idea on the particular brand of weirdness I'm experiencing here?
Thanks!
|
|
|
|
|
It's a long shot (and probably teaching grandmother to suck eggs...)...but is your callback interface dual or a dispinterface? You want it to be a dispinterface.
|
|
|
|
|
Hey, I'll take any kind of shot I can get at this point.
I haven't been nose deep in ActiveX controls since the late 90s (and I've slept since then). Any notion as to why the difference in interface types could give me trouble?
And thanks for the help, man!
Chris
|
|
|
|
|
Christopher Duncan wrote: Any notion as to why the difference in interface types could give me trouble?
Yeah, a dual interface implements methods/properties as 'normal' COM methods (i.e. C++ virtual methods) as well as through IDispatch. This means that the callback interface you're implementing could be called through that rather than through Invoke.
A dispinterface only implements methods through IDispatch, so Invoke is the only entry point for callbacks.
|
|
|
|
|
I understand the difference between the types of interfaces, but what puzzles me is why it would work in an exe and not in a dll.
|
|
|
|
|
Different connection point container implementations? Only thing I can think of off the top of my head.
Another thing you could try (if you haven't already) is ATL's QI debugging (using _ATL_DEBUG_QI) - you can see what interfaces are being used. You can use DbgView[^] to trace the QI's when running outside Visual Studio.
|
|
|
|
|
Really appreciate you taking the time to help with this. Apologies for being a bit dense, but I want to make sure I understand...
I'm using the same code (it's a C++ class) and compiling that code in two different scenarios. In the test app exe, Invoke is called. In the toobar dll, it's not. Since it's the same code in both cases, does "different connection point implementations" refer to the connection point implementation of the control host?
That would seem a sensible neighborhood for the culprit since the control code is the same in both scenarios. So, if this is what you mean, would you have any suggestions on how I could debug the connection point implementations in these two scenarios?
Thanks again, man.
|
|
|
|
|
Christopher Duncan wrote: does "different connection point implementations" refer to the connection point implementation of the control host?
Yep
Christopher Duncan wrote: So, if this is what you mean, would you have any suggestions on how I could debug the connection point implementations in these two scenarios?
I guess the QueryInterface debugging would be a reasonable start. Also, if your interface is dual, make it a dispinterface - that should (he said!!) fix the problem.
|
|
|
|
|
Okay! First, I'll make it dispinterface, and if that doesn't do it I'll find some chickens to sacrifice.
Thanks!
|
|
|
|
|
Hi
I have created an ATL application exe type (jjjj) where I am using to create ATL component
it gives following error
QLView.obj : error LNK2005: _CLSID_MyAWrapper already defined in jjjj.obj
SQLView.obj : error LNK2005: _LIBID_Project1Lib already defined in jjjj.obj
SQLView.obj : error LNK2005: _IID_IMyAWrapper already defined in jjjj.obj
Debug/jjjj.exe : fatal error LNK1169: one or more multiply defined symbols found
what to do here.. please suggest ..
thanks
Lee
Leya
|
|
|
|
|
I think you'll find this is caused because you're #including initguid.h in more than one .cpp file.
Make sure that it's only included in *one* cpp file explicitly. If it's in stdafx.h, take it out of there and put it in a .cpp file instead.
|
|
|
|
|
Hi,
I want to build a ATL wrapper on c++ class and would like to use that component in VB.
I insert an ATL class MyAWrapper
I add one property HRESULT String([out, retval] BSTR *pVal)
one method HRESULT ClearString() to IMyAWrapper.
IDL file shows it then in MyAWrapper.h
i put following c++ code
class foo
{
std::string m_str;
public:
foo(std::string bar, bool baz) : m_str(bar)
{
if (baz) throw std::exception();
}
std::string get_string(void) { return m_str; }
void clear_string(void) { m_str = ""; }
};
in ATL class.. class MyAWrapper I addfollowing code.
STDMETHOD(FinalConstruct)(void)
{
try {
m_pfoo = new foo("test", false);
}
catch (...) {
return E_FAIL;
}
return S_OK;
}
in MyAWrapper.cpp, I put
STDMETHODIMP MyAWrapper::get_String(BSTR *pVal)
{
// TODO: Add your implementation code here
if (pVal == NULL)
return E_POINTER;
std::string sTemp = m_pfoo->get_string();
*pVal = ::SysAllocStringLen(sTemp.c_str(), sTemp.length());
return *pVal ? S_OK : E_OUTOFMEMORY;
}
STDMETHODIMP MyAWrapper::ClearString()
{
// TODO: Add your implementation code here
m_pfoo->clear_string();
}
Please tell me Are these stpes OK??
or
where to put c++ code?
I can not compile the project.. I get around 34 errors one of which is
'std' : is not a class or namespace name
Pleasew tell me how to have this perfectly function.
Thanks
Lee
Leya
|
|
|
|
|
yang__lee wrote: std' : is not a class or namespace name
Probably, you forgot to include header, string in this case.
Prasad
MS MVP - VC++
|
|
|
|
General 
News 
Suggestion 
Question 
Bug 
Answer 
Joke 
Praise 
Rant 
Admin 
Submit an article or tip
