[Box Backup] Bug fix - don't die on lstat failure

Ben Summers boxbackup@fluffy.co.uk
Mon, 13 Dec 2004 22:04:10 +0000

On 13 Dec 2004, at 21:58, Joe Krahn wrote:

> Ben Summers wrote:
>> On 13 Dec 2004, at 18:12, Joe Krahn wrote:
>>> I found why I had problems getting my home directory backed up.
>>> If bbackupd get a bad result of lstat(), it aborts.
>>> In my case, the problem is actually from having a DVD .ISO file.
>>> Files greater than 4GB in size fail with stat; they need stat64.
>>> So, stat64() support is worth adding.
>>> But, it should also not abort the whole backup run due to one bad 
>>> stat call. It's better to just skip the item for which fstat failed 
>>> by chaning BackupClientDirectoryRecord.cpp from:
>> Isn't enabling linux LFS supposed to make it call that function 
>> instead of lstat itself?
>> Failures reading a file don't cause the backup to abort, but it does 
>> assume (at the moment) that getting file information is going to 
>> work.
>> I'll make it more tolerant.
> Yes, I think you can just define _FILE_OFFSET_BITS=64 and 
> automatically get 64-bit versions of stat structs/functions. But, I 
> don't see this in the boxbackup sources.

Which version of Box Backup are you using? v0.09 should run a test to 
see if this does anything useful, and then put that define in all the 
makefiles if it does.

Can you run ./configure again, and see if it says

   Linux LFS: yes

in the tests at the beginning? If it says yes, you should see the 
required define on the compiler command line as make does it's magic.