Configuration macro mismatch leads to ungracious termination
Posted: Sun Mar 24, 2013 12:22 pm
Hi Jarl,
I'm not sure if I should file this as a bug or to the wish list (so it goes here).
In my environment I build RCF as a shared library with a specific set of configuration macros.
Later on I use RCF to link my executables against it.
I've noticed that in case I do not use exactly the same set of configuration macros in respect to serialization (for example omitting boost.serialization) this surely leads to a segmentation violation due to illegal use of a NULL pointer.
You mention this in your documentation but I would like to add some extra guards to prevent this from happening because I find it difficult to track the error back to the root cause.
One solution I could imagine of is to add some symbols to the library that would force the same set of configuration macros (either during the linking process or during initialization of RCF).
Do you have a best practice for this problem ?
Regards,
Volker
I'm not sure if I should file this as a bug or to the wish list (so it goes here).
In my environment I build RCF as a shared library with a specific set of configuration macros.
Later on I use RCF to link my executables against it.
I've noticed that in case I do not use exactly the same set of configuration macros in respect to serialization (for example omitting boost.serialization) this surely leads to a segmentation violation due to illegal use of a NULL pointer.
You mention this in your documentation but I would like to add some extra guards to prevent this from happening because I find it difficult to track the error back to the root cause.
One solution I could imagine of is to add some symbols to the library that would force the same set of configuration macros (either during the linking process or during initialization of RCF).
Do you have a best practice for this problem ?
Regards,
Volker