Folder.Close(true) causes folder corruption

[RustyShelf]

Hi Ben,

Please let me know if you are sick of me pestering you.

My problem is that when I delete some things from a folder, and then close it using the expunge flag:
[folder object].close(true);
The contents of my folder are corrupted (All of the messages remaining lose all of their content). I can reproduce this every time.

I am using mstor v0.9.6 with the following settings:
p.setProperty("mstor.mbox.useNioMapping", "false");
p.setProperty("mstor.meta.enabled", "false");

Any ideas?

Rusty

[Ben Fortuna]

Hi Rusty,

I've just tried the following test, and it seems to work ok:

        URLName url = new URLName("mstor:c:/temp/mstor_test");
        Properties p = new Properties();
        // disable metadata..
        p.setProperty("mstor.meta.enabled", "false");
        p.setProperty("mstor.mbox.useNioMapping", "false");
        Store store = new MStorStore(Session.getDefaultInstance(p), url);
        store.connect();

        Folder inbox = store.getDefaultFolder().getFolder("Inbox");
        inbox.open(Folder.READ_WRITE);
        Message message = inbox.getMessage(1);
        message.setFlag(Flags.Flag.DELETED, true);
        inbox.close(true);

Let me know how your code differs from this and I can try to replicate your problem.

regards,
ben

[Ben Fortuna]

Rusty,

I've just found a couple of things that could lead to data corruption in a multi-threaded environment (regarding the encoder/decoder and dateformat in the MboxFile). This may be the cause of the problem, but I can't guarantee it..

regards,
ben

[RustyShelf]

Hi Ben,

I wrote some test code (similar to what you have above, but with a folder of 50 or so messages, and more delete operations) and it appears to work fine.

So then I checked my code again, and I found an issue where a thread was being kicked off before another one had finished. This was causing two threads to both trying using mstor at once, possibly even for the same folder. I found one other small issue with the way I was handling messages as well. It looks like one (or both) of these was the cause of my corruption issues...

sorry for the confusion,

Rusty

[Ben Fortuna]

Hi Rusty,

No worries. I hope to improve concurrency support in mstor, or at least document where classes are not thread-safe. If you have any suggestions or requests for improving synchronisation let me know.

regards,
ben