Review Request 118612: Fix KDirWatch's unit tests on certain file system.

[David Faure]

-----------------------------------------------------------
This is an automatically generated e-mail. To reply, visit:
https://git.reviewboard.kde.org/r/118612/#review59939
-----------------------------------------------------------

Ship it!



src/lib/io/kdirwatch.cpp
<https://git.reviewboard.kde.org/r/118612/#comment41725>

    might be worth a short comment in the code (even if the git log has the longer comment).
    
    Otherwise any reader will wonder why we're adding a path that most of the time is already in there....


- David Faure


On June 8, 2014, 7:44 p.m., Matthew Dawson wrote:
> 
> -----------------------------------------------------------
> This is an automatically generated e-mail. To reply, visit:
> https://git.reviewboard.kde.org/r/118612/
> -----------------------------------------------------------
> 
> (Updated June 8, 2014, 7:44 p.m.)
> 
> 
> Review request for KDE Frameworks.
> 
> 
> Repository: kcoreaddons
> 
> 
> Description
> -------
> 
> When using QFileSystemWatcher, try to avoid having deleted files stop being watched.
> 
> When a file is deleted, QFileSystemWatcher does not explicitly signal this fact,
> instead emitting a generic changed signal.  However, it will stop watching the file
> in question, at least on Linux with the inotify backend.  Normally this is ok, but
> if a file is quickly recreated and the inode number is unchanged, KDirWatch assumes
> the file is still monitored and no further changes to the file will occur.
> 
> Instead, KDirWatch will now always try to re-add the file to QFileSystemWatcher.
> It doesn't bother checking if the operation succeeded, as in most cases it will
> fail as the file is still watched.  It can only assume the file is now being
> watched at this point.
> 
> 
> Fix unit tests for KDirWatch when TMPDIR's filesystem reuses inode numbers.
> 
> When the filesystem TMPDIR points to (by default /tmp) reuses inode numbers,
> the testDeleteAndRecreateFile test will fail as neither the stat nor the
> QFileSystemWatcher backends can detect the deletion/recreation as two discrete
> events.  They only detect the fact the file changed, and emit the dirty signal.
> Fix the test suite to check for this case, and allow the test to proceed.
> 
> 
> Diffs
> -----
> 
>   autotests/kdirwatch_unittest.cpp c3dba4391b8ebcbedeaa53d51f2a170aa5695ccb 
>   src/lib/io/kdirwatch.cpp 7da1fdba8cd1a2e6a7d5bf9cef999d50b44ebf99 
> 
> Diff: https://git.reviewboard.kde.org/r/118612/diff/
> 
> 
> Testing
> -------
> 
> Unit tests all pass on all four backends (FAM only tested against gamin)
> 
> 
> Thanks,
> 
> Matthew Dawson
> 
>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.kde.org/pipermail/kde-frameworks-devel/attachments/20140613/dd2d7066/attachment.html>