~drizzle-trunk/drizzle/development : changes

~drizzle-trunk/drizzle/development » Changes from revision 910.4.22

Rev	Summary	Authors	Date
910.4.22	merge mainline merge mainline	Stewart Smith	15 years ago
910.4.21	merge mainline merge mainline	Stewart Smith	15 years ago
910.4.20	merge merge	Stewart Smith	15 years ago
910.4.19	merge other sparc fixes merge other sparc fixes	Stewart Smith	15 years ago
910.4.18	fix alignment of variables (incl key cache). fixes SIGBUS on... fix alignment of variables (incl key cache). fixes SIGBUS on SPARC64	Stewart Smith	15 years ago
910.4.17	merge mainline merge mainline	Stewart Smith	15 years ago
910.4.16	truncated variable is error (Work In Progress) test fixes truncated variable is error (Work In Progress) test fixes	Stewart Smith	15 years ago
910.4.15	is possible to get deadlock in a drizzleslap test case, we j... is possible to get deadlock in a drizzleslap test case, we just remove the --commit instead	Stewart Smith	15 years ago
910.4.14	Add --commit=100 to several drizzleslap runs in drizzleslap.... Add --commit=100 to several drizzleslap runs in drizzleslap.test to dramatically reduce the number of fsync()s when running the test. This takes the run time from ~393 seconds down to ~138seconds for me on the T1000 with a single disk. Not as fast as with libeatmydata (~42 seconds) - but a lot better. This really helps on Solaris as I was getting timeouts (15mins) before this patch.	Stewart Smith	15 years ago
910.4.13	batch up more INSERTs into transactions to help tests run qu... batch up more INSERTs into transactions to help tests run quicker.	Stewart Smith	15 years ago
910.4.12	dramatically speed up derived test. dramatically speed up derived test. Wrap 10,000 row load in transaction. Shaves 2 seconds off test for libeatmydata runs. Shaves about 335 seconds off test for normal runs (non-libeatmydata. i.e. fsync is real)	Stewart Smith	15 years ago
910.4.11	fix default value Item generation. (spotted on Solaris) fix default value Item generation. (spotted on Solaris) creating the Item just references the string we're passing in, not copy for own storage. This means we were returning an Item from default_value_item() with references to (now freed) objects on the stack. SunStudio on Solaris wipes this for us and makes things fail... a lot.	Stewart Smith	15 years ago
910.4.10	fix system variables for correct endian architectures. fix system variables for correct endian architectures. - remove sys_var_long (replace with 32 or 64 bit) - union { uint32_t, uint64_t } isn't endian portable for setting uint64 and reading from uint32 - start to have out of range variables throw error instead of truncating and setting	Stewart Smith	15 years ago
910.4.9	fix size_t variables on platforms where sizeof(size_t)=4 fix size_t variables on platforms where sizeof(size_t)=4	Stewart Smith	15 years ago
910.4.8	merge mainline merge mainline	Stewart Smith	15 years ago
910.4.7	ensure there is enough space allocated to String for string ... ensure there is enough space allocated to String for string representation of date/time types. The API to the date_to_str functions is COMPLETE ASS anyway and should be taken out and shot.	Stewart Smith	16 years ago
910.4.6	Fix format string for error message. thread_id is int64_t. Fix format string for error message. thread_id is int64_t. Causing SIGSEGV on Linux/SPARC and Linux/PowerPC. Oddly enough, not on Solaris/SPARC.	Stewart Smith	16 years ago
910.4.5	remove out of date comment remove out of date comment	Stewart Smith	16 years ago
910.4.4	max_sort_length should be size_t everywhere. Causing numerou... max_sort_length should be size_t everywhere. Causing numerous failures on SPARC and PowerPC due to strang value being retrieved in filesort. Basically, anything with filesort fails without this patch.	Stewart Smith	16 years ago
910.4.3	merge latest dev work merge latest dev work	Stewart Smith	16 years ago

Older »