[ https://issues.apache.org/jira/browse/JCI-67?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13085339#comment-13085339 ]
Sebb commented on JCI-67: ------------------------- Safer would be the following, as it checks the path is actually a directory: {code} final File parent = file.getParentFile(); if (!parent.mkdirs() && !parent.isDirectory()) { throw new IOException("could not create" + parent); } } {code} > Dubious use of mkdirs() return code > ----------------------------------- > > Key: JCI-67 > URL: https://issues.apache.org/jira/browse/JCI-67 > Project: Commons JCI > Issue Type: Bug > Reporter: Sebb > Priority: Minor > > FileRestoreStore.java uses mkdirs() as follows: > {code} > final File parent = file.getParentFile(); > if (!parent.exists()) { > if (!parent.mkdirs()) { > throw new IOException("could not create" + parent); > } > } > {code} > Now mkdirs() returns true *only* if the method actually created the > directories; it's theoretically possible for the directory to be created in > the window between the exists() and mkdirs() invocations. > Also, the initial exists() call is redundant, because that's what mkdirs() > does anyway (in the RI implementation, at least). > I suggest the following instead: > {code} > final File parent = file.getParentFile(); > if (!parent.mkdirs() && !parent.exists()) { > throw new IOException("could not create" + parent); > } > } > {code} > If mkdirs() returns false, the code then checks to see if the directory > exists, so the throws clause will only be invoked if the parent really cannot > be created. > The same code also appears in AbstractTestCase and > FilesystemAlterationMonitorTestCase. -- This message is automatically generated by JIRA. For more information on JIRA, see: http://www.atlassian.com/software/jira