提交 356d8626 authored 作者: Thomas Mueller's avatar Thomas Mueller

Avoid deadlock while closing

上级 626f165d
......@@ -38,6 +38,33 @@ and there is a header after each chunk)
header:
H:3,...
Format:
Current store header:
H:3,blockSize:4096,chunk:10,creationTime:1391059293945,format:1,lastMapId:15,rootChunk:8192,version:10,fletcher:5d2a9623
used:
chunk,creationTime,format,(formatRead,)lastMapId,rootChunk,version
(blockSize not used)
fletcher
map.10 = test
name.lobData = 10
chunk: store live only if there is garbage!
Plan: (hex encoded values, H:2, rootChunk -> block, creationTime -> created, lastMapId -> map
{H:2,block:2,blockSize:1000,chunk:a,created:143e19856f9,format:1,map:b,version:a,fletcher:5d2a9623}\n
{block:2,blocks:a0,chunk:a,max:2030,pages:100,root:a020,time:103}\n
{<chunk>}\n .... {<storeHeader>}\n
map.a = test
name.lobData = a
Chunk: (id -> chunk, start -> block, length -> blocks, pageCount -> pages,
pageCountLive -> livePages, maxLength -> max, maxLengthLive -> liveMax,
metaRootPos -> root (offset))
+, if different: maxLive:1030,pagesLive:30
compression: support multiple algorithms
TODO:
Documentation
......@@ -121,6 +148,8 @@ MVStore:
specially for large pages (when using the StreamStore)
- StreamStore: split blocks similar to rsync crypto, where the split is made
"if the sum of the past 8196 bytes divides by 4096 with zero remainder"
- Compression: try using a bloom filter before trying to match
- DataType: change to reading and writing arrays, not individual entries
*/
......@@ -667,7 +696,7 @@ public class MVStore {
try {
fileStore.writeFully(pos, buffer);
} catch (IllegalStateException e) {
close();
closeImmediately();
throw e;
}
}
......@@ -1959,6 +1988,10 @@ public class MVStore {
return;
}
backgroundWriterThread = null;
if (Thread.currentThread() == t) {
// within the thread itself - can not join
return;
}
synchronized (t.sync) {
t.sync.notifyAll();
}
......
Markdown 格式
0%
您添加了 0 到此讨论。请谨慎行事。
请先完成此评论的编辑!
注册 或者 后发表评论