Crash with AssertionError while backing up to exFAT partition about rdiff-backup

[ericzolf]

The truth is that rdiff-backup has issues with other (more or less) case insensitive file systems, see #11, so we should improve the way we detect and treat those file systems in the most possible safe way.

This issue will probably get closed together with #11 but I keep it separate for now as there is a slight chance that the solution will be different.

[ottok]

Hello @rik-shaw !

The code base has been rewritten for Python 3 and there will be no further releases for the 1.x series in Python 2 anymore. Could you please help our efforts and test again with the latest development version (https://github.com/rdiff-backup/rdiff-backup#installing-the-latest-development-version) if you can still reproduce this and provide us with a test case (steps on how to reproduce the issue) that applies for latest development version?

With your help, if we can reproduce this problem still with the 2.x code base if might be fixed it in upcoming 2.x series releases.

[ericzolf]

The issue is still there in 2.0.3.

[ericzolf]

That's a funny one and has nothing to do with rdiff-backup, I can reproduce the issue independently of it, on an exfat volume:

1. touch A; rm a; stat A -> no error!!! stats are shown
2. touch A; rm A; stat A -> No such file error
3. touch A; rm a; sleep 0.9; stat A -> still no error, stats are shown
4. touch A; rm a; sleep 1; stat A -> No such file error, at last after one full second!

It sounds to me like there is a timer of 1 second in the exFAT driver before the names are synced. I tried to sync in-between but that didn't help. We could introduce a sleep of 1 second also in the code, but that's very ugly. I couldn't find anything in the man-pages to change the behaviour of exfat. I'll plug a bug towards the driver and see what the developers have to say. Perhaps they'll have an idea how to workaround/overcome.

[ikus060]

What if you call fsync?

…

On Tue., Jun. 30, 2020, 1:08 p.m. Eric L., ***@***.***> wrote: That's a funny one and has nothing to do with rdiff-backup, I can reproduce the issue independently of it, on an exfat volume: 1. touch A; rm a; stat A -> no error!!! stats are shown 2. touch A; rm A; stat A -> No such file error 3. touch A; rm a; sleep 0.9; stat A -> still no error, stats are shown 4. touch A; rm a; sleep 1; stat A -> No such file error, at last after one full second! It sounds to me like there is a timer of 1 second in the exFAT driver before the names are synced. I tried to sync in-between but that didn't help. We could introduce a sleep of 1 second also in the code, but that's very ugly. I couldn't find anything in the man-pages to change the behaviour of exfat. I'll plug a bug towards the driver and see what the developers have to say. Perhaps they'll have an idea how to workaround/overcome. — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#38 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAHA5I6ZB3ASRUQUUBI4LFLRZILXHANCNFSM4GFSHBRA> .

[ericzolf]

That's the sync I was describing, no effect sadly.

[ericzolf]

OK, I thought about it and even if I could easily fix the test with a sleep 1, I can't guarantee that the rest of the code will work properly under all circumstances on a file system with such a flaw in its implementation, so I will improve the error handling and close the bug with it, but not try to fix the test.

[brainchild0]

I'll plug a bug towards the driver and see what the developers have to say.

Is this issue still open upstream? Would you provide a link?

Is the arter97/exfat-linux project relevant to resolving full support for exFAT?