[aspectc-user] calls to template members not matched

[Olaf Spinczyk]

Hi,

Panu Bloigu wrote:
> Hi,
> 
>> On the other hand --real-instances sometimes leads to parsing problems
> 
> I just tested the --real-instances mode and it seems to fail with the code
> shown in my previous post. However, if I remove references to iostream, the
> code copiles fine and the joinpoint is matched! I confirmed this by doing
> the printing with stdio's printf. So, to me it seems like ac++ is going
> through all the included headers and fails with iostream, beacuse of it
> being overwhelmingly complex. Might this assumption be correct?
> 
> Also, is there any way to predict, even approximately, what kind of code
> might be too much for --real_instances? Can I include any STL headers, or
> just some, and how can I find out (besides by trial) which ones would
> compile fine?
> 
> Panu.

at least for Linux I can say that you can *include* the STL headers.
Problems may occur only when you actually *use* the templates from the
STL. If this is not too much scattered over your code base, you could
wrap points where you use the STL with #ifndef __puma ... #endif.

It might also help you to know that there are two ac++ test programs
that use the STL in the real-instances mode: one uses <string>, the
other uses <set>. It might be that case that the containers work but the
streams are problematic.

Does that work for you?

Olaf