It occurred to me – in the page-search code, we copy the page's write generation, then search the page.
It occurs to me we should probably have a memory barrier instruction after we copy the write generation, otherwise the compiler might re-order that read after we've actually searched the page? (Or make page->write_gen volatile?)
--keith
- related to
-
WT-1 placeholder WT-1
- Closed
-
WT-2 What does metadata look like?
- Closed
-
WT-3 What file formats are required?
- Closed
-
WT-68 Design and implement dump format
- Closed
-
WT-103 --enable-python requires --enable-shared
- Closed
-
WT-175 race in forced page eviction
- Closed
-
WT-179 the cache isn't empty after verifying WiredTiger.wt
- Closed
-
WT-310 overflow blocks may be accessed after deletion
- Closed
-
WT-322 table and lsm object statistics
- Closed
-
WT-403 Minor changes to fix uri table display. WT-397
- Closed
-
WT-441 Allow LSM trees to discard the btree handle from the active chunk.
- Closed