Richard Melville wrote:
On 29 October 2015 at 16:13, Bruce Dubbs <[email protected]> wrote:
Pierre Labastie wrote:
Hi,
While running OpenJDK tests, I realized that one test was failing because
it
expected to find "uptime" in /usr/bin, but we install it in /bin, as we do
for
all procps executables (One could complain that tests shouldn't hard code
paths, but it is not my point). This raises the question of what files
should
go into /bin (and /sbin).
I cannot see how "uptime" could be needed early in the boot process, or in
case /usr is not mountable. Same for "free", "w", more arguably for
"watch",
"top" and so on.
OTOH, the FHS is clear:
------
/bin contains commands that may be used by both the system administrator
and
by users, but which are required when no other filesystems are mounted
(e.g.
in single user mode). It may also contain commands which are used
indirectly
by scripts.
------
My question is therefore: should we try to move some of the executables in
/bin to /usr/bin? If yes, then the second question is "which executables".
I
can propose a list (only tomorrow actually; I'll have other things to do
this
afternoon and tonight)
We have the same problem with eudev where one of the regression tests hard
codes /usr/bin/test and we have it in /bin (where it needs to be).
It's not unreasonable for uptime to be in /usr/bin, but as it is I only see
107 executables in /bin and 11 of those are symbolic links.
I don't think it's necessary to move files just because some upstream dev
hard coded a path into a test. Just use a sed to fix it.
It's not a random act by an "upstream dev"; as far as I can see all
the above exist in /usr/bin in most, if not all, the mainstream
distros.
It is a random act to hard code a path in the tests. Most distros are
now using systemd which we do not use. systemd does not support
separate /usr partitions and we do.
-- Bruce
--
http://lists.linuxfromscratch.org/listinfo/lfs-dev
FAQ: http://www.linuxfromscratch.org/faq/
Unsubscribe: See the above information page
