ATL / WTL / STL

11-Jul-09 7:13

I am putting together a local COM server, which gets linked to a legacy static library. My server passes on notifications from this library functions to the client via connection point sinks (server has the Fire_xxx methods).

My problem is, the library author has freely used CreateThread()s in his functions, from which he triggers the notifications that I need to pass on to the client. As you guessed, this throws 8001010e HR. To get around this issue I used to register the interface ptr in GIT and getting it in the context of the thread. This trick wont work here, as with just the base interface, I am not able to touch all the Fire_XXX() functions!

How to get access to these? TIA.

Stuart Dootson11-Jul-09 7:30

11-Jul-09 7:30

Why not decouple the notifications from the Fire_XXX functions? Use an event, or similar to indicate to the COM thread that a notification has occurred (you can use some data to show what notification has occurred plus any parameters). Use a kernel wait in the COM thread to wait for the event to be signalled. When the event is signalled, call the appropriate Fire_XXX function.

Java, Basic, who cares - it's all a bunch of tree-hugging hippy cr*p

Kannan Ramanathan11-Jul-09 7:41

11-Jul-09 7:41

I thought of this...but, I am not sure how to apply this on my case..lemme explain: The server implements, lets say, "IServer" with a method "Run". I use "IServerEvents" to fire events in to connected sinks.

From IServer::Run(), I call one of the library functions, which does a CreateThread() and returns. So, my Run() returns a S_OK and things are all fine so far. But, when the thread started by the library function finishes its work and wants to notify the client, it calls a callback function my server had registered with it in the start. I call Fire_XXX() in this callback, which is not working as it is in the context of another thread...

Since my IServer::Run() has ended long time back, which COM context do I use?

Stuart Dootson11-Jul-09 7:52

11-Jul-09 7:52

What threading model does your local server use? Is it possible that using a multithreading model might enable your original design to work?

Java, Basic, who cares - it's all a bunch of tree-hugging hippy cr*p

Kannan Ramanathan11-Jul-09 18:23

11-Jul-09 18:23

I use MTA; Still the same issue. BTW, since the thread created by library function doesn't initialize COM, in my server callback I do a CoInitializeEx to COINIT_MULTITHREADED. From this callback when I try calling Fire_XXX (in the original server object, which is passed via a context param), I still get 8001010e.

Will creating a CWorkerThread/IWorkerThreadClient based thread automatically assume its parent thread COM context?

Stuart Dootson11-Jul-09 23:37

11-Jul-09 23:37

I presume you have managed to successfully call one of the Fire_XXX functions in tests or something? It's not that they just don't work?

Kannan Ramanathan wrote:
Will creating a CWorkerThread/IWorkerThreadClient based thread automatically assume its parent thread COM context?

I suspect you're more likely to know than me - I think you've donw a lot more COM than me. My comfort zone is mostly in-proc servers.

Java, Basic, who cares - it's all a bunch of tree-hugging hippy cr*p

Kannan Ramanathan11-Jul-09 23:43

11-Jul-09 23:43

Thanks Stuart for the reply.

Yes, Fire_xxx() events work correctly as long as they are called in the correct context.

BTW, I am trying out my good old GIT solution -- only this time I am going to store the sink interfaces in GIT and store the cookie in my vector...I am currently putting together a custom IConnectionPointImpl<> to store/retrieve things properly with out breaking anything;

Will update you how it goes...

Stuart Dootson11-Jul-09 23:47

11-Jul-09 23:47

The other suggestion I was going to make (but didn't get around to) was to start a long-running thread and use the event idea I first proposed within that thread.

Java, Basic, who cares - it's all a bunch of tree-hugging hippy cr*p

Kannan Ramanathan12-Jul-09 1:57

12-Jul-09 1:57

Custom implementation of IConnectionPointImpl<> to make use of GIT worked like a charm! Smile | :)

I'd seriously advise anyone stuck in a similar situation to try out this trick -- do let me know if you need my implementation (I think it shouldnt be a problem sharing my code, as it doesnt have any project specific details in it!)

Thanks Stuart for your replies.

Stuart Dootson12-Jul-09 5:48

12-Jul-09 5:48

Hint, hint - article -hint, hint. Smile | :)

Java, Basic, who cares - it's all a bunch of tree-hugging hippy cr*p

VC6 problem: locale loc(locale::classic(), new myfac)

Member 348023215-Jul-09 0:51

Member 3480232

15-Jul-09 0:51

Much larger than just a hint. I, for one, have been fighting problems like this for a while now and would love to look at a different view than my own for a change. Thumbs Up | :thumbsup:

John R. Shaw9-Jul-09 20:51

John R. Shaw

9-Jul-09 20:51

Yes I am still supporting VC6 in this project and it has to be cross platform.

Buried in the VC6 MSDN Library under “locale::facet” is the line
“locale loc(locale::classic(), new myfac);”.

Which is correct for latter versions of the MS STL, but there is no locale constructor that takes a locale and a facet defined in the VC6 STL version of locale.

My unsatisfactory workaround is:

#ifdef _ADDFAC // Dinkumware macro
    std::locale myLoc = std::_ADDFAC(std::locale::classic(), new my_codecvt);
#else
    std::locale myLoc = std::locale(std::locale::classic(), new my_codecvt);
#endif

Do you have a better one?

INTP
"Program testing can be used to show the presence of bugs, but never to show their absence."Edsger Dijkstra

Re: VC6 problem: locale loc(locale::classic(), new myfac)

Stuart Dootson9-Jul-09 22:01

How to hide Standard Toolbar button in IE

9-Jul-09 22:01

I'm sure this is teaching grandma to suck eggs, but the only thing I'd suggest is to encapsulate that in another function, so you only need to do it once, i.e:

template<class Facet>
std::locale AddFacet(std::locale const& l, const Facet* f)
{
#ifdef _ADDFAC // Dinkumware macro
    return std::_ADDFAC(std::locale::classic(), new my_codecvt);
#else
    return std::locale(std::locale::classic(), new my_codecvt);
#endif
}

Alternatively, use this:

#ifndef _ADDFAC // Dinkumware macro
#define _ADDFAC (std::locale)
#endif

and use _ADDFAC for all platforms.

Java, Basic, who cares - it's all a bunch of tree-hugging hippy cr*p

am 20098-Jul-09 22:40

am 2009

8-Jul-09 22:40

Hello,

I want to hide standard toolbar button of IE. i can hide IE std .toolbar , but affetcs my othere toolbar also. So is there any way to hide only Standard Toolbar button of IE.

thanks
Am

32bit client talking to a 64bit local server

Kannan Ramanathan4-Jul-09 3:58

4-Jul-09 3:58

Is it possible to have a 32bit client initialize/invoke methods in a 64bit local COM server? If yes, how do I set up the Proxy/stub DLLs? The target OS I have in mind is Vista64 or Win7-64 (so the 32bit client will be running from with in WOW64).

Also, can I go ahead with ATL 8 for implementing the above scenario? Any examples available online? Thanks.

Kannan Ramanathan4-Jul-09 4:00

4-Jul-09 4:00

One more req I have to worry abt: The 64bit local server should be able to implement connection points (32bit client being the sink)...is this possible? Thanks.

Stuart Dootson4-Jul-09 4:50

4-Jul-09 4:50

This thread[^] documents the opposite scenario (64-bit client, 32-bit server) - does that enlighten the situation at all?

And this thread[^] indicates that 32-bit client and 64-bit client should be OK.

Java, Basic, who cares - it's all a bunch of tree-hugging hippy cr*p

Kannan Ramanathan4-Jul-09 5:32

4-Jul-09 5:32

Thanks Stuart for the pointers. Looks like all I need to do is to build/register both 32bit and 64bit proxy/stubs. Is that right?

In this case, can I also make the 32bit code act as a sink? Thanks.

Stuart Dootson4-Jul-09 5:35