git-svn-id: file:///svn/toku/tokudb@44972 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44970 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44967 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44962 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44961 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44960 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44959 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44958 c7de825b-a66e-492c-adef-691d508d4ae1

#refs 5133 merge undo-do algorithm changes to main. now, we read a row from the source db AND gather up the leaf entrie's provisional data in the cursor callback, which makes it atomic with respect to commit and abort on that DB. had to fix a few tests after changing some leaf entry signatures


git-svn-id: file:///svn/toku/tokudb@44949 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44940 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44937 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44919 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44902 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44899 c7de825b-a66e-492c-adef-691d508d4ae1

close[t:5134] le_cursor api cleaned up, followed up on a TODO to get rid of the DB parameter from is_key_right (and improved its name too) by using a fake_db in the le_cursor plus the new toku_ft_get_descriptor() api to grab a descriptor reference. refactored some tests to get it all compiling. LE and hot indexer tests pass


git-svn-id: file:///svn/toku/tokudb@44895 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44893 c7de825b-a66e-492c-adef-691d508d4ae1

[t:5064] Closes #5064 Move most of the work of creating xids (including malloc) out of the txn_manager_lock

git-svn-id: file:///svn/toku/tokudb@44892 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44890 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44885 c7de825b-a66e-492c-adef-691d508d4ae1

[t:5132] cast pthread_t through uintptr_t in order to print it in a way that's portable between linux and osx.  closes #5132

git-svn-id: file:///svn/toku/tokudb@44882 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44874 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44864 c7de825b-a66e-492c-adef-691d508d4ae1

also adds csv and tsv output formats (and some work on gnuplot)

git-svn-id: file:///svn/toku/tokudb@44863 c7de825b-a66e-492c-adef-691d508d4ae1

close[t:5128] updating a descriptor now has two parts: user API that passes a FT and descriptor, and a low layer API that takes those two plus an explicit descriptor. we use the low layer one for deserialization code that wants to upgrade a descriptor for an FT that is not fully opened yet (and thus whose FT does not have a valid cachefile yet). this seems to be a clean enough solution, since it helps some upgrade code do non-standard stuff.

most importantly, loader-stress-test3 no longer fails by segfaulting on ft->cf->fd


git-svn-id: file:///svn/toku/tokudb@44839 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44837 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44836 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44835 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44834 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44829 c7de825b-a66e-492c-adef-691d508d4ae1

[t:4838] the locktree now opens its own ft handle when created to ensure that the ft stays in memory so long as the lock tree is in memory. we use an on_create and on_destroy callback to keep the txn/ft logic out of the locktree. the api for toku_ltm_get_lt is a little hairy now. a better solution may come in the future. for now, the original failing test passes.


git-svn-id: file:///svn/toku/tokudb@44819 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44814 c7de825b-a66e-492c-adef-691d508d4ae1

close[t:5124] use a cleaner way of access an FT_HANDLE's descriptor pointers at the ydb layer. comment the ft API, too.


git-svn-id: file:///svn/toku/tokudb@44811 c7de825b-a66e-492c-adef-691d508d4ae1

[t:5113] carefully document a non-obvious contract requirement for change descriptor. clients may ONLY update a FT's cmp descriptor if it's the first operation done after open, before any reads or writes. further, remove the test - it's user error.


git-svn-id: file:///svn/toku/tokudb@44804 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44780 c7de825b-a66e-492c-adef-691d508d4ae1

[t:5113] spruce up the default parameters and put a reader-writer lock around the global variable used in the assertion, so we know for certain we're not just racing to read THAT variable


git-svn-id: file:///svn/toku/tokudb@44778 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44777 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44776 c7de825b-a66e-492c-adef-691d508d4ae1

[t:5112] removed the fd param from update descriptor, improved the name, removed some unnecessary return codes from functions that cannot fail

git-svn-id: file:///svn/toku/tokudb@44771 c7de825b-a66e-492c-adef-691d508d4ae1

git-svn-id: file:///svn/toku/tokudb@44769 c7de825b-a66e-492c-adef-691d508d4ae1

[t:5110] Closes #5110 Skip taking the txn_manager lock for read-only transactions during note_abort/commit

git-svn-id: file:///svn/toku/tokudb@44758 c7de825b-a66e-492c-adef-691d508d4ae1