784888 - Test for bug 674742 to prevent dataloss regression due to folder compact

Status: RESOLVED INVALID

(MailNews Core :: Backend, defect)

Description

[Hiroyuki Ikezoe (:hiro)]

Attachment: The test

Bug 674742 causes dataloss. It should be tested.

This test needs a stub of nsIFile (the patch for bug 784870).

Comment 1

[Hiroyuki Ikezoe (:hiro)]

The test should also check the content of the folder equals after compaction. I will post a new test with the approach that the target file/directory make read-only.

Comment 2

[Hiroyuki Ikezoe (:hiro)]

Changing dependency.

The test depend on the patch for bug 784870 since the result of nsIMsgFolderCompactor.compact should be checked in the test.

Comment 3

[Hiroyuki Ikezoe (:hiro)]

Attachment: Try to make the parent folder of the compaction target file read-only (incomplete)

This is a try.

Unfortunately making the target file read-only does not work well at all. It can work no errors! 

This patch tries to make the parent directory of the target read-only. It gets NS_ERROR_FILE_ACCESS_DENIED at CreateUnique in nsFolderCompactState::Init().
Sadly the failure result does not return, so this test hangs...

I will open a new bug for this hang.

Comment 4

[Hiroyuki Ikezoe (:hiro)]

Attachment: Try to make the parent folder of the compaction target file read-only (incomplete)

I am sorry that I posted the patch before qrefresh.

Comment 5

[Hiroyuki Ikezoe (:hiro)]

(In reply to Hiroyuki Ikezoe (:hiro) from comment #3)
> This patch tries to make the parent directory of the target read-only. It
> gets NS_ERROR_FILE_ACCESS_DENIED at CreateUnique in
> nsFolderCompactState::Init().
> Sadly the failure result does not return, so this test hangs...
> 
> I will open a new bug for this hang.

Filed bug 785274.

Comment 6

[Hiroyuki Ikezoe (:hiro)]

Attachment: Another test with nsIFileStub

Yay! Now the test with nsIFileStub works fine without any patch against mozilla-central.

Comment 7

[Hiroyuki Ikezoe (:hiro)]

Mark, now I'd recommend the second test even if it needs the stub for nsIFile. I will maintain the stub code as possible.

Comment 8

[Hiroyuki Ikezoe (:hiro)]

I can write a new test code after bug 786936 is closed.

Comment 9

[Wayne Mery (:wsmwk)]

an example of where our compact tests are substandard

Comment 10

[Wayne Mery (:wsmwk)]

Can we call this and bug 784870 obsolete?

Comment 11

[Ben Campbell]

Yes, I think so... but they are larger issues: our testing of outside-the-app I/O failures (disk, network, whatever) is pretty awful at the moment.
I think it's worth another more comprehensive attempt, so I just wrote up Bug 1897667.