Open brianrmurphy opened 1 year ago
Which version of dupd are you running? Looking at the source, that particular message is probably from 1.6 or earlier.
I suggest first try with the latest released version (1.7.1).
In general though, such an error might be caused by the file contents/size changing while the scan is running.
Just compiled from git master branch. Here's the stack at the point where the error message is about to print.
Thread 3 "dupd" hit Breakpoint 2, read_entry_bytes (entry=0x7fffd894a642, filesize=4763056, path=0x7ffff780eed0 "/mnt/4t/w530-brm/Desktop/admin/oldcopy/usr/lib/locale/locale-archive", output=0x7ffd2c5b3090 "arch?aut\341", <incomplete sequence \371>, bytes=0, skip=65536, bytes_read=0x7ffff780edd8) at src/utils.c:196 196 printf("error: requested zero bytes from [%s] (skip=%" PRIu64 ")\n", (gdb) where
path=0x7ffff780eed0 "/mnt/4t/w530-brm/Desktop/admin/oldcopy/usr/lib/locale/locale-archive", output=0x7ffd2c5b3090 "arch?aut\341", <incomplete sequence \371>, bytes=0, skip=65536, bytes_read=0x7ffff780edd8) at src/utils.c:196
entry=0x7fffd894a642, path=0x7ffff780eed0 "/mnt/4t/w530-brm/Desktop/admin/oldcopy/usr/lib/locale/locale-archive") at src/sizelist.c:438
src/sizelist.c:605
pthread_create.c:477
../sysdeps/unix/sysv/linux/x86_64/clone.S:95 (gdb)
This is after Processed 117207/161500 files, so log files and output are crazy big, but here's the last few lines of output (with -v -v -v): Processed 117204/161500 (0 duplicates of size 27697584) Processed 117205/161500 (0 duplicates of size 9976956) Processed 117206/161500 (0 duplicates of size 12763568) Processed 117207/161500 (0 duplicates of size 18555312)
and trace file: 1073930 A RB 65536 275828834 1073930 A RB 65536 275894370 1073931 A RB 65536 275959906 1073932 A RB 65536 276025442
On Wed, Dec 21, 2022 at 11:46 AM Jyri J. Virkki @.***> wrote:
Which version of dupd are you running? Looking at the source, that particular message is probably from 1.6 or earlier.
I suggest first try with the latest released version (1.7.1).
In general though, such an error might be caused by the file contents/size changing while the scan is running.
— Reply to this email directly, view it on GitHub https://github.com/jvirkki/dupd/issues/45#issuecomment-1362014241, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANWAPKDDFCUB4SAU332WWTWONNARANCNFSM6AAAAAATF6SHQA . You are receiving this because you authored the thread.Message ID: @.***>
Thanks, I see.
The development version (master branch) might be in an unstable state as it is work in progress, although I'm not aware of specific bugs in it so this is helpful. You could still try the released version (1.7.1) to see if it hits the same issue or not.
Seems like the file size and block layout on disk is just so that it triggers a bug in the next block size computation.
A few debugging commands may give more useful info:
Disk layout for the file as dupd sees it:
dupd info --x-extents /mnt/4t/w530-brm/Desktop/admin/oldcopy/usr/lib/locale/locale-archive
There's also an option to print more debug output only for a given size to avoid overwhelming amount of output:
dupd scan --debug-size 4763056
(Those seem to be the size and path of the file causing the crash from the stack trace, but change accordingly if not.)
Same issue in branch 1.7_maintenance. Is that the same as the 1.7.1 release?
On Wed, Dec 21, 2022 at 10:44 PM Jyri J. Virkki @.***> wrote:
Thanks, I see.
The development version (master branch) might be in an unstable state as it is work in progress, although I'm not aware of specific bugs in it so this is helpful. You could still try the released version (1.7.1) to see if it hits the same issue or not.
Seems like the file size and block layout on disk is just so that it triggers a bug in the next block size computation.
A few debugging commands may give more useful info:
Disk layout for the file as dupd sees it:
dupd info --x-extents /mnt/4t/w530-brm/Desktop/admin/oldcopy/usr/lib/locale/locale-archive
There's also an option to print more debug output only for a given size to avoid overwhelming amount of output:
dupd scan --debug-size 4763056
(Those seem to be the size and path of the file causing the crash from the stack trace, but change accordingly if not.)
— Reply to this email directly, view it on GitHub https://github.com/jvirkki/dupd/issues/45#issuecomment-1362476480, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANWAPOSRTGFNGB56V6EJIDWOP2GPANCNFSM6AAAAAATF6SHQA . You are receiving this because you authored the thread.Message ID: @.***>
With the debug-size flag, I see the following. Perhaps something leaps out at you from it, but it seems a bit insufficient to me.
Files scanned: 1560000 ...
On Thu, Dec 22, 2022 at 1:10 PM Brian R. Murphy @.***> wrote:
Same issue in branch 1.7_maintenance. Is that the same as the 1.7.1 release?
On Wed, Dec 21, 2022 at 10:44 PM Jyri J. Virkki @.***> wrote:
Thanks, I see.
The development version (master branch) might be in an unstable state as it is work in progress, although I'm not aware of specific bugs in it so this is helpful. You could still try the released version (1.7.1) to see if it hits the same issue or not.
Seems like the file size and block layout on disk is just so that it triggers a bug in the next block size computation.
A few debugging commands may give more useful info:
Disk layout for the file as dupd sees it:
dupd info --x-extents /mnt/4t/w530-brm/Desktop/admin/oldcopy/usr/lib/locale/locale-archive
There's also an option to print more debug output only for a given size to avoid overwhelming amount of output:
dupd scan --debug-size 4763056
(Those seem to be the size and path of the file causing the crash from the stack trace, but change accordingly if not.)
— Reply to this email directly, view it on GitHub https://github.com/jvirkki/dupd/issues/45#issuecomment-1362476480, or unsubscribe https://github.com/notifications/unsubscribe-auth/AANWAPOSRTGFNGB56V6EJIDWOP2GPANCNFSM6AAAAAATF6SHQA . You are receiving this because you authored the thread.Message ID: @.***>
Please include output of
dupd info --x-extents /mnt/4t/w530-brm/Desktop/admin/oldcopy/usr/lib/locale/locale-archive
it should be same as in the size debug output, but good to confirm.
I'm running on a pretty big set of files (4M+ files, about 2TB), and I get this error partway through.
Progress line looks like: Sets : 117207/ 161500 65138564K ( 64302K/s) 0q 1%b 3075f 1013 s error: requested zero bytes from [omitted_filename] (skip=65536)
The mentioned file exists, has data, and is readable. I am already trying to reproduce in gdb for more info, but suggestions/ideas could be useful.