[
https://issues.apache.org/jira/browse/LUCENE-6957?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15085360#comment-15085360
]
Adrien Grand commented on LUCENE-6957:
--------------------------------------
I'm confused about the latest patch:
{code}
+ IndexOutput out = null;
+ try {
+ while (true) {
+ out = first.createTempOutput(prefix, suffix, context);
+ String name = out.getName();
+ toDelete.add(name);
+ if (slowFileExists(second, name)) {
+ out.close();
+ } else {
+ success = true;
+ break;
+ }
+ }
+ } finally {
+ if (success) {
+ IOUtils.deleteFiles(first, toDelete);
+ } else {
+ IOUtils.closeWhileHandlingException(out);
+ IOUtils.deleteFilesIgnoringExceptions(first, toDelete);
+ }
+ }
+
+ return out;
{code}
Won't this always return an output on a deleted file since we always call
{{toDelete.add(name)}} regardless of whether the file exists in the other
directory? Should we make toDelete a HashSet and call {{toDelete.remove(name)}}
in the else block?
> NRTCachingDirectory is missing createTempOutput
> -----------------------------------------------
>
> Key: LUCENE-6957
> URL: https://issues.apache.org/jira/browse/LUCENE-6957
> Project: Lucene - Core
> Issue Type: Bug
> Reporter: Michael McCandless
> Assignee: Michael McCandless
> Fix For: Trunk
>
> Attachments: LUCENE-6957.patch, LUCENE-6957.patch
>
>
> It's broken now because it simply delegates to the wrapped dir now,
> which can create an output that already exists in the ram dir cache.
> This bug only affects trunk (it's never been released).
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
