Windows Build Broken

[Rodolfo Ochoa]

Error 21 error C2079: 'tb' uses undefined struct 'zorba::DateTime::getLocalTime::__timeb64' \src\zorbatypes\datetime\datetimetype.cpp 328 1 zorba_simplestore

Error 22 error C3861: '_ftime64_s': identifier not found \src\zorbatypes\datetime\datetimetype.cpp 329 1 zorba_simplestore

[Rodolfo Ochoa (rodolfo-ochoa)]

test_hashmaps is still failing because of using maps with zstring as key, this is unacceptable for C++11 standard, and the last conversation through email lead to nothing, so, I'm putting std::string instead as key, after all, the key is what is being test in this case.

[Paul J. Lucas (paul-lucas)]

The failure-to-compile in test_hashmaps has nothing to do with the C++11 standard. The only reason it doesn't work is because Markos didn't include the necessary header to define std::hash() for zstring which is util/hash/hash.h.

[Chris Hillery (ceejatec)]

Paul - Rodolfo was able to get things going with your change; thanks.

However, in order to do this, he had to force the compilation of src/util/hash/* by removing the check at the bottom of src/util/CMakeLists.txt. Currently this directory is omitted if ZORBA_HAVE_UNORDERED_MAP or ZORBA_HAVE_UNORDERED_SET are true, and both are true on Windows.

This doesn't seem quite right; I assume that the checks for ZORBA_HAVE_UNORDERED_MAP/HASH are there for a reason. Can you tell us what the Right Thing To Do(tm) is here, in addition to adding the #include you mentioned?

1. Add src/util/hash/* to the project all the time?
2. Skip adding test_hashmap to the project if ZORBA_HAVE_UNORDERED_MAP/HASH are set?
3. Something else?

Assign this back to Rodolfo with the instructions.

[Chris Hillery (ceejatec)]

Bumping the priority high since this is preventing any development on Windows.

[Paul J. Lucas (paul-lucas)]

OK, there was one small problem that was my fault: hash.cpp (and only hash.cpp) should always be compiled. I've fixed that in the CMakeLists.txt files.

However, as I've been saying repeatedly: Markos' test_hashmaps.cpp file is testing *hashmaps* and not zstring, so there's no reason why he *must* use zstring there -- std::string will do just fine. I've replaced zstring there with std::string.

Assigning back to Rodolfo. (I have no easy way to test this on Windows.)