Lines Matching refs:to

3 # The author disclaims copyright to this source code.  In place of
14 # more database pages than it has to, stuff like that).
37 #           fsync() calls on the journal file, no need to set nRec
77 # INSERT causes 2 db pages to be written - the root-page of "abc"
92 # to be added to the root page of abc. 4 pages in total
93 # are written to the db file - the two leaf pages, the root
101 # the root-node, which is close to empty, and two leaf pages, both of 
113 # This insert should use the quick-balance trick to add a third leaf
114 # to the b-tree used to store table abc. It should only be necessary to
115 # write to 3 pages to do this: the change-counter, the root-page and
134 # The following INSERT updates 2 pages and requires 4 calls to fsync():
137 #   2) The journal file (to sync the page data),
138 #   3) The journal file (to sync the journal file header),
146 # Set the device-characteristic mask to include the SQLITE_IOCAP_ATOMIC,
147 # then do another INSERT similar to the one in io-2.2. This should
207 # modifies a single database page and also appends a page to the file.
208 # Internally, this case is handled differently to the one above. The
213 # bytes 28..31 and so when a page is added to the database, page 1
235 } {1 {unable to open database file}}
246 # is created during execution of the COMMIT statement, so we have to
247 # use the same technique to check that it is created as in the above 
271   } {1 {unable to open database file}}
392     # File size might be 1 due to the hack to work around ticket #3260.
417     # have been no calls to fsync(). The file is probably about 30KB.
424     # The COMMIT requires a single fsync() - to the database file.
436 # than 4, calls to fsync(). The fsync() calls are on:
439 #   2) The journal file (to sync the page data),
460 # be set to 0xFFFFFFFF before the first journal sync. The nRec field
480 # written to the database, even though the sync-mode is "full".
516   # difficult to predict the exact size of the file as reported by 
521     # abc). Because the cache_size is set to 10, this must have required
523     # to the journal file after the cache-spill, then the size of the
618   # cache size to become too small to hold the entire db in memory.
636   # to be flushed. Which is a bug.