Closed HorlogeSkynet closed 4 years ago
If we stick with df
, we should probably go with df -k -P
, since that's about as much as we can do following the original UNIX specification (see here) and convert the 1024-byte blocks into a human readable size. That should give us the most compatibility with anything *NIX.
I think we will/should ; It's fast, relevant and pretty-well supported.
So agreed :100:
Although, as we will lose -l
, it's definitely relevant to work this out during #55, with the upcoming ability to specify mount-points.
Note to myself : Maybe it's time to rework the way we invoke
ps
inProcesses
at the same time. Current best candidate :ps -eo comm
. Header line will have to be ignored. --> + remove the behavior when "run as root" note in README. Fixed inaef1cd0bc89182005b088e8a127d6aa5648a67aa
.
EDIT :
Note to myself (2) : Using
-k
will also fix the wholeDisk
usage as the current-B MB
parameter is misused.
This will be indirectly fixed by #67 ; closing here.
Describe the bug On Alpine Linux, for instance, the
Disk
entry is left undetected. See https://github.com/HorlogeSkynet/archey4/commit/d676b0b2d58c4a4f0e6beefcfbd5456f0165a6ec#diff-e6076e0f8c9ab0586c8909ec8d6832beR67.Additional context When fixing this (let's say with a simpler
df -P -B M
call and a proper columns parsing), we've to think about an optimized approach, as while checking for BTRFS mount-points, the seconddf
call would also fail... Maybe this fix could be thought over while reworkingDisk
for #55.