Re: [PATCH v7 1/2] e2fsck: Correct ext4 dates generated by old kernels.

2013-12-07 Thread Theodore Ts'o
On Sat, Dec 07, 2013 at 09:58:17PM -0500, David Turner wrote: > > We can and should change time_to_string to take an unsigned 64-bit > > type; it's an internal interface to debugfs. > > Shouldn't this be a signed 64-bit type, since we have to support times > before 1970? That depends on whether

Re: [PATCH v7 1/2] e2fsck: Correct ext4 dates generated by old kernels.

2013-12-07 Thread David Turner
On Sat, 2013-12-07 at 19:53 -0500, Theodore Ts'o wrote: > On Sat, Dec 07, 2013 at 03:02:40PM -0500, David Turner wrote: > > > > However, as Andreas notes, "we want to verify .. that "debugfs -R > > 'stat testfile'" decodes the times correctly." Unfortunately, it > > does not, and it is not trivia

Re: [PATCH v7 1/2] e2fsck: Correct ext4 dates generated by old kernels.

2013-12-07 Thread Theodore Ts'o
On Sat, Dec 07, 2013 at 03:02:40PM -0500, David Turner wrote: > > However, as Andreas notes, "we want to verify .. that "debugfs -R > 'stat testfile'" decodes the times correctly." Unfortunately, it > does not, and it is not trivial to fix. debugfs uses an internal > function time_to_string(__u3

Re: [PATCH v7 1/2] e2fsck: Correct ext4 dates generated by old kernels.

2013-12-07 Thread Andreas Dilger
I suspect that any 32-bit systems running at that time will have been updated to have 64-bit time_t or otherwise have windowed the 32-bit time_t to have a new starting epoch. So I'm willing to punt on decoding the 64-bit value correctly to libc and just assign our time to the system time_t. C