Hi,
The GNU tar documentation states about the --checkpoint-action format
string:
‘%{N}*’
Pad output with spaces to the Nth column. If the ‘{N}’ part is
omitted, the current screen width is assumed.
This does not work right in multibyte locales, when non-ASCII characters
are present on the line, before the padding.
Examples:
$ LC_ALL=ru_RU.UTF-8 tar cf /tmp/all.tar --checkpoint-action="echo=Hit
[%t]%{50}*|" coreutils-*.tar.gz
...
tar: Hit [Чт 30 окт 2025 13:20:18] |
...
The vertical bar is in column 45, not column 50.
$ LC_ALL=ru_RU.UTF-8 tar cf /tmp/all.tar --checkpoint-action="echo=Hit
[%t]%{50}*|" coreutils-*.tar.gz
...
tar: Hit [2025年10月30日 13時21分23秒] |
...
The vertical bar is in column 38, not column 50.
$ tar --version
tar (GNU tar) 1.35
Copyright (C) 2023 Free Software Foundation, Inc.
To me, it looks like the cause is that in tar/src/checkpoint.c, the
return value of function format_checkpoint_string counts bytes, not
columns. To count columns, it should use the Gnulib module 'mbswidth'.
Bruno
