On 27/12/2023 20:52, Jim Meyering wrote:
Thanks. Nice improvement.
I compared --help and --version output, which showed no difference
other than the intended ones.
Thanks for the review. Pushed.
make check passed, modulo this unrelated failure on Fedora 38:
numfmt.pl: test lcl-dbl-to-human-1: stdout mismatch, comparing
lcl-dbl-to-human-1.1 (expected) and lcl-dbl-to-human-1.O (actual)
*** lcl-dbl-to-human-1.1 Wed Dec 27 12:47:48 2023
--- lcl-dbl-to-human-1.O Wed Dec 27 12:47:48 2023
***************
*** 1 ****
! 1,1K
--- 1 ----
! 1,1k
Oh interesting.
I suspect that's due to glibc-all-langpacks not being installed on your system.
That would mean a regular space was used as the thousands grouping char,
and thus the test was run for you, and not for me.
I see a similar skipping issue in tests/sort/sort-h-thousands-sep.sh,
and in fact that was hiding a sort bug introduced in v9.1 !
Hopefully the 3 patches attached addresses these issues.
Paul are you Ok with the change to sort in the last patch?