…nnecting to config server
Requested to revert code.
Fixing util/paths.h fixed the "wrong assert" problem, so the changes to db/matcher.cpp and db/btree.cpp are not needed now.
We need to replace the assert definition that Boost defines with our version, otherwise code that includes this header won't work.
Make btree.cpp use the correct assert while headers are being investigated.
The code in matcher.cpp detects when it is using the wrong assert() macro in release builds and that code was triggering, preventing mongod from passing unit tests and making BuildBot useless. This fix fixes matcher.cpp only and is just to get the builds going again. Now looking for which header file did this to us. More to come ...
FlushViewOfFile() will return error code 33 (ERROR_LOCK_VIOLATION) in some cases, but this is a "transient" error, and just retrying repeatedly will (supposedly) always work. This changes the code to retry up to one million times and for up to 60 seconds if it continues to get ERROR_LOCK_VIOLATION.
…CK_EXCEPTION is to be used only with the boost unit test framework, not as a general tool to catch exceptions. Signed-off-by: Eric Milkie <email@example.com>
This patch does the following: 1.) Remove pcre.h dependencies in the C++ client, and remove some other unnecessary dependencies. 2.) Clean up how we build the client from the client source tarball, so it's more like how we build it from the git repo / full source tarball. 3.) Fix up our "scons" so you only have to write "scons ." to build all of our binaries, the binary archive (zip or tgz) and client source archive (zip or tgz). 4.) Fix up SCons aliases "core", "tools", and "all". 5.) Let user specify the name of the client tarball via a SCons command line switch. Resolves SERVER-4231, SERVER-5255.
…wed to be nested for reading. so changing the test to match that
…when writing to admin you need to lock both in this change we simply W lock the whole system on admin writes. this should make writing to auth work. however there may be better solutions. we may be able to treat admin db as non-nested on writes period.
tr1 contains boost's shared pointer and thus the symbol names conflict unless you explicitly scope them. "using boost::shared_ptr" isn't sufficient.
…en appending values in explain.cpp
…ure consistency after a write
…adedObject() to MatchDetails::hasLoadedRecord().