Open darealshinji opened 1 year ago
Using 1.26.3 on Fedora-39 - copying 200GB of data files between disks has already taken 24hrs, and it's getting slower. It takes less than 2 hours using Fedora's standard "Files". It's running from a terminal showing these errors repeatedly:
Glib-gio-critical 00:34:49.964 gfileinfo created without stsndard::sort-order
"" Gfileinfofile ../gio/gfileinfo.c line2062 (g_file_info_get_symlink-target):should not be reached
Also some errors like this:
CAJA-IS-BOOKMARK (bookmark) failed
G_IS_FILE (file) failed
G_IS_OBJECT (object) failed
This could relate to GVFS or metadata handling issues, and is also reported with Nautilus and Nemo. This level of slow could also be slow disk drives or connections. A USB 2 hard drive case for instance would never get over 29MB/sec.
How much faster would this job run with cp from the command line on your system?
This is not related to a slow disk, I observe the same issue when coping/deleting big directories in Caja or Nautilus, in comparison with doing the same operation in the terminal. I already get used to that for moving big amounts of data I need to open mc
in the terminal instead of using Caja/Nautilus.
For Nautilus it is well described in the following issue:
https://gitlab.gnome.org/GNOME/nautilus/-/issues/1904
From the discussion it seems the deleting may be optimized by using unlinkat
syscall, like rm -r
do it.
However it is true this is probably more GVFS issue than issue of Nautilus/Caja/Nemo...
I gave up after about 36hours copying, as the number of hours remaining stayed the same. This was because the bitrate halved roughly every 8 hours. It was down to 1.5MB/s when I gave up. Zeno's paradox is real! It said there was 200GB to copy and "Disk Usage Analyser" thought there was 200GB too, but in fact there was only 150 according to Fedora "Files" and that eventually proved to be the case. The difference may have been in '.' files.
I've just used Bash
cp -rf /mnt/old/chris/* ~
it took 40 minutes and it was 150GB. So the problem is not related to disk speed, processor speed or memory size.
I'm now using Nemo 5.2.4 which seems to deal with large directories much better, especially when I just want to delete files. Disk speed can't really be an issue for me either, as I'm using SSD.
PS: if you want more details about progress and remaining copy time in a terminal you can try vcp or a similar tool for python whose name I forgot.
Expected behaviour
I delete a directory with over a thousand files inside (source tree, node_module stuff, chroot installation) and it shouldn't take much more than a second to do.
Actual behaviour
Takes several seconds. It's actually way faster to open a terminal and call
rm -rf
orcp -rf
. I assume the window isn't updated fast enough or something?Steps to reproduce the behaviour
Download something with more than a thousand files inside and try it out.
MATE general version
1.26.0
Package version
1.26.0-1
Linux Distribution
Ubuntu MATE 22.04
Link to bugreport of your Distribution (requirement)
https://bugs.launchpad.net/ubuntu/+source/caja/+bug/2001624