Open Hub
Open Hub

 Settings |  Report Duplicate

0
I Use This!
Inactive

Commits : Listings

Analyzed about 15 hours ago. based on code collected about 16 hours ago.
Aug 19, 2024 — Aug 19, 2025
Commit Message Contributor Files Modified Lines Added Lines Removed Code Location Date
KNF
Keith Bostic
More... about 16 years ago
Add STOC LRU page discard function. Fix function names (__wt_cache_db -> __wt_cache_)
Keith Bostic
More... about 16 years ago
Fix function names (__wt_cache_db -> __wt_cache_)
Keith Bostic
More... about 16 years ago
typo
Keith Bostic
More... about 16 years ago
Use WT_ERR, there's an error label we can use, and the destroy function should work even if there's no ienv.
Keith Bostic
More... about 16 years ago
use "tags", not "flags" makes more sense.
Keith Bostic
More... about 16 years ago
WT_PSTOC_NOT_SET not used.
Keith Bostic
More... about 16 years ago
typo, forgot WT_STOC_FOREACH's arguments
Keith Bostic
More... about 16 years ago
Add WT_STOC_FOREACH macro.
Keith Bostic
More... about 16 years ago
typo: missed a "goto err;".
Keith Bostic
More... about 16 years ago
Add WT_ERR, WT_RET and WT_TRET macros.
Keith Bostic
More... about 16 years ago
Now that we require an ENV, there's no need for a set of globals, we can put that information into the IENV structure. Remove TOC single-threaded flag, use IENV one, instead.
Keith Bostic
More... about 16 years ago
Kill the thread if we're NOT single-threaded.
Keith Bostic
More... about 16 years ago
Gcc doesn't like single quotes inside #if 0 blocks.
Keith Bostic
More... about 16 years ago
Db handles now map one-to-one to STOC handles, and are created when the Db handle is opened/closed; the STOC has the work queue now, there's no global work queue. Add STOC->running field, so can shut a server down when the Db handle is closed. The cache is now passed a STOC, not a Db handle; rename __wt_cache_db_XXX functions to be __wt_cache_XXX; remove code to walk a global cache, it's all STOC local now. Clarify close & sync functions, which goes in Db code, which in Btree code. Remove Btree code to fork off a server thread for internal database pages. Add code to sleep the STOC if it's not busy for long enough. Add statistics to the STOC. Remove cachesize-verify routine, no longer necessary (there's still no code to re-size the cache, though). Pages no longer require a file ID (the STOC is per-file), or a server ID (which they'd never used until now). Move file-handle reference counts into the OS layer. Make OS sleep function regularize its arguments.
Keith Bostic
More... about 16 years ago
Add DB-local sync function.
Keith Bostic
More... about 16 years ago
Pin the root page for each database at open. Move the cache into the server thread structure. Make the Env the only standalone structure, everything else (specifically creating the DB and TOC structures) is a method call off the Env. Change the work q to be identified by a server ID # + a TOC
Keith Bostic
More... about 16 years ago
Pin the root page for each database at open. Move the cache into the server thread structure. Make the Env the only standalone structure, everything else (specifically creating the DB and TOC structures) is a method call off the Env. Change the work q to be identified by a server ID # + a TOC
Keith Bostic
More... about 16 years ago
typo
Keith Bostic
More... over 16 years ago
The DBT callback routine has to get the key, even if the key isn't changed; pass the key to the DBT return function in all cases.
Keith Bostic
More... over 16 years ago
Add support for a callback function out of the DBT.
Keith Bostic
More... over 16 years ago
checkpoint
Keith Bostic
More... over 16 years ago
Lint.
Keith Bostic
More... over 16 years ago
Add a "build the in-memory information" flag to __wt_bt_page_in, then replace uses of __wt_cache_db_in with calls to __wt_bt_page_in with that flag off. (In general, stat, verify and dump don't need the in-memory structures.)
Keith Bostic
More... over 16 years ago
Restore the "number of levels in the tree" statistic, it's still useful.
Keith Bostic
More... over 16 years ago
Remove the level from the on-disk page header -- the page's type plus encoding if an item references an internal or leaf page in the item type is sufficient. Remove the addr from the in-memory index structure, we can use the INDX ditem field to retrieve it from the page. The dump code doesn't use the in-memory structures, so don't build them (use the cache_db_in/out API, not the bt_page_in/out API). Change __wt_bt_first_offp() to not overwrite anything, instead pass back the address, and if the page is a leaf or internal page: quit passing WT_ITEM_OFFP structures when recursively walking the tree (the stat and verification code), just pass an addr and a flag if the page is a leaf or internal page.
Keith Bostic
More... over 16 years ago
Smoke-test program & scripts.
Keith Bostic
More... over 16 years ago
Accidentally deleted a line when doing 8B-to-4B array casting.
Keith Bostic
More... over 16 years ago
Fix some comments.
Keith Bostic
More... over 16 years ago
Solaris & Linux both pad structures that contain a u_int64_t followed by a u_int32_t (either before or after), to an 8-byte boundary. The WT_ITEM_OFFP structure needs to be 24 bytes, it's the chunk we store on internal pages. Casts works for now, but we might have to use bit masks if we find a compiler where casts don't work. Solaris: SunOS coolthreads 5.11 snv_61-xb_15 sun4v sparc SUNW,Sun-Fire-T200 cc: Sun C 5.8 Patch 121015-02 2006/03/29 Linux: Linux pcls3 2.6.27.7-9-default #1 SMP 2008-12-04 18:10:04 +0100 x86_64 x86_64 x86_64 GNU/Linux gcc (SUSE Linux) 4.3.2 [gcc-4_3-branch revision 141291]
Keith Bostic
More... over 16 years ago