cp (GNU coreutils) 8.32
CentOS Linux release 7.4.1708 (Core)
kernel 3.10.0-693.17.1.el7.x86_64
I have identified an issue with "cp" where, if more than 2 TiB of data is
being copied, and progress reporting is on, after 2 TiB has been copied the
time remaining overflows.
The output goes like this:
728 files copied so far... 2.0 TiB /
4.9 TiB
[============================> ]
41.0 %
Copying at 206.2 MiB/s (about 4h 25m 44s remaining)
...127TE.RDM/A018_A007_0127H2.RDC/A018_A007_0127H2_004.R3D 1.6 GiB /
3.8 GiB
[=============================> ]
42.1 %
728 files copied so far... 2.0 TiB /
4.9 TiB
[============================> ]
41.0 %
Copying at 203.1 MiB/s (about 4294967286h 4294967261m 4294967249s remaining)
...127TE.RDM/A018_A007_0127H2.RDC/A018_A007_0127H2_004.R3D 1.7 GiB /
3.8 GiB
[===============================> ]
44.7 %
This is "cp -argu <sourcedir> <destdir>". The source tree contains more
than 2TiB worth of data.
I believe the issue is in src/copy.c where (on line 355) an INT is used to
store "cur_size".
int cur_size = g_iTotalWritten + *total_n_read / 1024;
When dealing with filesizes and capacities, the largest possible type
should be used, which I believe is a unsigned long long (or __uint64_t).
Note that I also think that usec_elapsed and sec_elapsed should be larger
types than int and double to prevent overflows.
