Sears Russell
c2fa9913b1
Added partial implementation of range tracking for pinning arbitrary
...
regions. Currently, tracks regions, and is O(m log(n)) for each
operation, where m is the number of transition points per range, and n
is the number of transition points in the system. A transition point is
a place in memory where the pincount changes.
2007-01-24 01:03:52 +00:00
Sears Russell
f6ccaadc35
Updated slotted api to match fixed api; cleaned up dead code.
2007-01-22 20:55:25 +00:00
Sears Russell
9b49787931
typo in getSetRaw()
2007-01-22 20:54:04 +00:00
Sears Russell
346fe97161
Fixed race (wasn't grabbing lock); typo in getInstantSet
2007-01-22 20:53:26 +00:00
Sears Russell
3aa83e7a48
Added link to test.c from main API page, and doxygenified the test.c comments.
2006-12-18 01:12:30 +00:00
Eric Brewer
fcb0a33da9
New directory for testing application code
2006-12-18 00:48:18 +00:00
Sears Russell
f522aab6ad
Typo (misspelled logfile name)
2006-12-14 02:53:55 +00:00
Sears Russell
495db15775
Documentation fix
2006-12-14 02:53:14 +00:00
Sears Russell
4e4461a193
check.h contains the work "end", which is #defined by the compenstaion stuff; moved check.h to be the first include.
2006-11-22 05:05:58 +00:00
Sears Russell
5b56370980
Fixed append. (It wasn't calling updateEOF, as it should have been. Also, updateEOF assumed start_pos was zero)
2006-11-22 05:04:48 +00:00
Sears Russell
3c3ccd9df7
Uncommented the file and memory tests.
2006-11-22 05:03:32 +00:00
Sears Russell
802eeb33e7
New non-blocking implementation, with backpressure. Still doesn't handle overlapping writes. (Multiple writes to the same region is OK though.)
2006-11-21 06:50:12 +00:00
Sears Russell
4520f672ac
The test sometimes calls append, append, truncate, where the first append terminates after the append, and returns EDOM. Check for this, and allow the test to pass.
2006-11-21 06:49:16 +00:00
Sears Russell
1dbf1145ca
Update to detect changes in file size (to support multiple file handles...)
2006-11-21 06:47:35 +00:00
Sears Russell
a06077d12d
Conditionally compile benchmarks
2006-11-21 06:45:22 +00:00
Sears Russell
a1a53d26f5
Updated front page of API documentation.
2006-11-13 21:49:57 +00:00
Sears Russell
6ba05cd619
Fixes BerkeleyDB support.
2006-11-11 00:39:41 +00:00
Sears Russell
75b0de74cd
Fixes inline warning on 32bit intel
2006-11-11 00:36:48 +00:00
Sears Russell
cb90e015a1
Fixed printf warnings
2006-11-10 19:32:44 +00:00
Sears Russell
c7591e8311
Better conditional compilation; handles missing libraries gracefully.
2006-11-10 11:18:50 +00:00
Sears Russell
329b5b26b9
fixed compile warnings
2006-11-10 11:18:24 +00:00
Sears Russell
b8d0ccce79
Modularized the build; made berkeleydb and confuse optional. No longer build as many things by default.
2006-11-09 23:00:50 +00:00
Sears Russell
499fdc9dfa
Removed FILE_OFFSET_BITS, as it is now set by ./configure.
2006-11-09 23:00:09 +00:00
Sears Russell
e0cdcebf2f
Added pointer to web page.
2006-11-09 22:58:58 +00:00
Sears Russell
9294f89277
Fix for fallback posix_memalign implementation. (Inserted missing
...
casts)
2006-11-08 07:22:42 +00:00
Sears Russell
f458e4fcd8
Fixed compilation directions.
2006-11-08 02:25:25 +00:00
Sears Russell
450261f3c5
Fixed 32bit bug.
2006-11-07 22:37:05 +00:00
Sears Russell
6efa862485
Fixed 64->32bit cast warnings.
2006-11-07 22:36:46 +00:00
Sears Russell
c86688884a
Updates for 32-bit x86. There are still real 32-bit bugs being emitted as compiler warnings.
2006-11-07 20:36:59 +00:00
Sears Russell
8481e9cdd6
non_blocking is closer to working (it now flushes writes out in the background). This required some api changes. It still passes unit tests, but it's losing data on close()...
2006-10-28 03:33:02 +00:00
Sears Russell
bb9e339823
changed to deal with new LinkedList api.
2006-10-28 03:31:27 +00:00
Sears Russell
aa445bdfe5
slightly more general purpose; deals with longs instead of ints.
2006-10-28 03:30:45 +00:00
Sears Russell
faf55911df
Added non-blocking io implementation. It doesn't actually work yet, but it passes the unit tests. ;)
2006-10-26 05:48:30 +00:00
Sears Russell
10b77729f5
Simple (blocking, poor truncation performance) file handle implementation.
2006-10-18 00:57:36 +00:00
Sears Russell
f2e7a6b3a8
Fixed bug (assert was calling a handle method with an outstanding read buffer).
2006-10-17 03:28:24 +00:00
Sears Russell
b948e72518
Removed the rwl* code; no longer includes assert.h
2006-10-17 03:27:17 +00:00
Sears Russell
32347b12e4
Added a "debug" handle dumps a trace of requests passed to it, and passes the requests through to another handle implementation.
2006-10-17 03:26:01 +00:00
Sears Russell
45ef627c49
Updated to build check_io.c
2006-10-13 09:03:32 +00:00
Sears Russell
80a4148543
First cut at a new io API, with an in-memory reference implementation + test case.
2006-10-13 04:39:03 +00:00
Sears Russell
e961a2bdb8
Fixed compile warning.
2006-10-07 20:05:54 +00:00
Sears Russell
e81bf95553
Added a 10 minute timeout.
2006-10-07 00:26:18 +00:00
Sears Russell
8bb5192a41
Updated coverage stuff so that it completes even if some unit tests
...
fail, and so that it only runs tests in test/lladd.
2006-10-06 23:46:53 +00:00
Sears Russell
1575d97a6b
Fixed recovery bug. TregionDealloc was being called during redo(), causing an assert to fail. Without the
...
assert, would have gone into infinite recursion.
2006-10-05 22:08:42 +00:00
Sears Russell
71f51d11bc
Fixed a performance bug that was causing logical operations to unnecessarily pin pages.
2006-10-05 22:05:50 +00:00
Sears Russell
a3e3ad2453
Added a 20 minute timeout to prevent deadlocks and other problems from blocking the test.
2006-10-05 01:00:10 +00:00
Sears Russell
8092fcb5a5
Made test longer again. (The group commit signal strangeness was preventing it from terminating...)
2006-10-05 00:52:41 +00:00
Sears Russell
f0612e7353
Documentation Fix
2006-10-05 00:51:24 +00:00
Sears Russell
3054256c6e
Documentation fix
2006-10-05 00:51:01 +00:00
Sears Russell
6a9d81c78f
Removed unneeded #includes.
2006-10-05 00:50:46 +00:00
Sears Russell
836d2d097d
Fixed initialization bugs on pendingCommits and syncLogCount. Group commit used to hang with 100% CPU if a signal interrputed it. Now it ignores the signal. (Don't know why it's receiving a singal, so it warns when this happens...)
2006-10-05 00:50:21 +00:00
