attachment set

Thu, 21 Jun 2012 15:50:16 GMT

attachment → ser.patch

component changed; owner set

lukester_null@… — Wed, 27 Jun 2012 10:08:02 GMT

owner set to Robert Ramey
component None → serialization

anonymous — Thu, 12 Jul 2012 16:10:22 GMT

hmmm - the sun compiler fails the trunk test suite. Since you're using this and the serialization library, maybe you could run the serialization library tests on your system and track down the cause of these failures. I suspect it's just a standard io library quirk but there's nothing I can do about it on my own. I believe that a little sleuthing would be rewarded with a couple of changes which would result if the sun compiler passing (almost) all tests.

Robert Ramey

lukester_null@… — Thu, 12 Jul 2012 16:50:01 GMT

Actually looking at it again, looks like it would throw up some warnings with VC++ and it does indeed:

basic_iarchive.cpp
..\..\..\libs\serialization\src\basic_iarchive.cpp(345) : warning C4267: 'argument' : conversion from 'size_t' to 'const unsigned int', possible loss of data
..\..\..\libs\serialization\src\basic_iarchive.cpp(380) : warning C4267: 'argument' : conversion from 'size_t' to 'const unsigned int', possible loss of data
..\..\..\libs\serialization\src\basic_iarchive.cpp(391) : warning C4267: 'argument' : conversion from 'size_t' to 'const unsigned int', possible loss of data
..\..\..\libs\serialization\src\basic_iarchive.cpp(470) : warning C4267: 'initializing' : conversion from 'size_t' to 'const unsigned int', possible loss of data
compile-c-c++ ..\..\..\bin.v2\libs\serialization\build\msvc-9.0\debug\address-model-64\threading-multi\basic_iserializer.obj
basic_iserializer.cpp
compile-c-c++ ..\..\..\bin.v2\libs\serialization\build\msvc-9.0\debug\address-model-64\threading-multi\basic_oarchive.obj
basic_oarchive.cpp
..\..\..\libs\serialization\src\basic_oarchive.cpp(279) : warning C4267: 'argument' : conversion from 'size_t' to 'const unsigned int', possible loss of data
..\..\..\libs\serialization\src\basic_oarchive.cpp(367) : warning C4267: 'argument' : conversion from 'size_t' to 'const unsigned int', possible loss of data
compile-c-c++ ..\..\..\bin.v2\libs\serialization\build\msvc-9.0\debug\address-model-64\threading-multi\basic_oserializer.obj
basic_oserializer.cpp

... and so on.

It also throws up a bench of warnings about unsafe functions. I normally don't see those as _SCL_SECURE_NO_WARNINGS and _CRT_SECURE_NO_DEPRECATE get set in my build environment (not via bjam).

So sorry about the confusion there. This was a patch I've been applying against boost for some time and I thought it was about time I mentioned it - but sadly forgot where it was generating the warnings in the first place!

If I get some time I'll have a look with SunStudio but hopefully the VC++ warnings can be removed in the meantime?

Best regards

Luke Elliott.

Robert Ramey — Sun, 15 Jul 2012 19:51:49 GMT

FYI, this is not so easy as it seems.

it presumes that the underlying type of object_id_type is unsigned int.

But in fact, it is uint_least32_t which may or may not be the same type as unsigned int.

I'm still looking into this.

Robert Ramey

Boost C++ Libraries: Ticket #7010: A few 32/64 bit warning fixes

attachment set

component changed; owner set