Re: [Boost-users] Serialization problem on 64-bit Windows release

Robert Ramey
This doesn't seem to be the case in your situation.
Anyway, thank you for the information. I'll keep trying to reproduce an issue in test lab. Sergey

Sergey Filippov
Robert Ramey
writes: This doesn't seem to be the case in your situation.
Anyway, thank you for the information. I'll keep trying to reproduce an issue in test lab.
Sergey
The issue is resolved. That has forced me to think in the beginning, that it is a mistake is specific only for 64-bit Windows, not so important. Neither incorrect optimization in the 64-bit compiler (as it was in the Sigurd's case) nor misoperation of the serialization library in multithreaded environment (as it is noticed by Robert) was a reason. The reason was very simple: in one of the classes the BOOST_CLASS_EXPORT macro was used before inclusion export.hpp, that led to the class registration code was not instantiated. It is a good lesson for me: the failure to find a time for writing a test framework has led to much greater losses of time. Thanks, Sigurd and Robert, for your time. Though your replies have not pointed to the real reason, they pushed me to think in another direction and eventually to resolve the problem. Sergey

Neither incorrect optimization in the 64-bit compiler (as it was in the Sigurd's case) nor misoperation of the serialization library in multithreaded environment (as it is noticed by Robert) was a reason.
Hmmm - I thought you were all the same person.
The reason was very simple: in one of the classes the BOOST_CLASS_EXPORT macro was used before inclusion export.hpp, that led to the class registration code was not instantiated.
Note that in the most current version of serialization, checked into the HEAD, instantiation of code has been improved by D. Abrahams. Its possible that in the future this would never have occured.
Thanks, Sigurd and Robert, for your time. Though your replies have not pointed to the real reason, they pushed me to think in another direction and eventually to resolve the problem.
You're welcome. I'm always happy to send someone off in the wrong direction.. Robert Ramey
participants (2)
-
Robert Ramey
-
Sergey Filippov