Skip to content

file.exists() on device files

6 messages · Benjamin Tyner, Rolf Turner, Henrik Bengtsson +1 more

Original
Benjamin Tyner
# 
Hi,

On my linux machine (Ubuntu, and also tested on RHEL), I am curious to 
know what might be causing file.exists (and also normalizePath) to not 
see the final device file here:

    > list.files("/dev/fd", full.names = TRUE)
    [1] "/dev/fd/0" "/dev/fd/1" "/dev/fd/2" "/dev/fd/3"
    > file.exists(list.files("/dev/fd", full.names = TRUE))
    [1]  TRUE  TRUE  TRUE FALSE
    > normalizePath(list.files("/dev/fd", full.names = TRUE))
    [1] "/dev/pts/2" "/dev/pts/2" "/dev/pts/2" "/dev/fd/3"
    Warning message:
    In normalizePath(list.files("/dev/fd", full.names = TRUE)) :
      path[4]="/dev/fd/3": No such file or directory

    > sessionInfo()
    R version 3.2.5 (2016-04-14)
    Platform: x86_64-pc-linux-gnu (64-bit)
    Running under: Ubuntu 16.04.1 LTS

    locale:
     [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
     [3] LC_TIME=en_US.UTF-8        LC_COLLATE=en_US.UTF-8
     [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
     [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
     [9] LC_ADDRESS=C               LC_TELEPHONE=C
    [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C

    attached base packages:
    [1] stats     graphics  grDevices utils     datasets  methods base

Regards
Ben
Rolf Turner
#
On 11/01/17 23:12, Benjamin Tyner wrote:
(a) Exactly the same thing happens to me (I am also running Ubuntu 16.04).

(b) Things get a bit confused because /dev/fd is actually a symbolic link:
(c) But then doing
Gives the same result as before:
(d) It turns out that the four "files" in /proc/self/fd are again
symbolic links:
(e) But now do it again!!!
Different number; 7154 rather than 7150.

(f) The name "/proc" would seem to imply that this has something to do 
with processes; the directories "7150", "7154" etc. are being created 
and removed on the fly, as a result of some process (presumably the "ls"
process) starting and finishing.

I have no insight into what is being effected here, or what is really 
going on "deep down", but the foregoing is some sort of "explanation".
By the time file.exists() is invoked, the ls process called by 
list.files() has finished and the associated directory (e.g. "7150", 
"7154", ...) has ceased to be.

What you do with this "explanation" is up to you.  My advice would be to 
forget about it and go to the pub! :-)

cheers,

Rolf Turner
Henrik Bengtsson
#
On Wed, Jan 11, 2017 at 3:56 PM, Rolf Turner <r.turner at auckland.ac.nz> wrote:
FYI, the /proc is there because Unix has something called the "proc
filesystem (procfs; https://en.wikipedia.org/wiki/Procfs) is a special
filesystem in Unix-like operating systems that presents information
about processes and other system information in a hierarchical
file-like structure".  For instance, you can query the uptime of the
machine by reading from /proc/uptime:

$ cat /proc/uptime
332826.96 661438.10

$ cat /proc/uptime
332871.40 661568.50


You can get all IDs (PIDs) of all processes currently running:

$ ls /proc/ | grep -E '^[0-9]+$'

and for each process you there are multiple attributes mapped as
files, e.g. if I start R as:

$ R --args -e "message('hello there')"

then I can query that process as:

$ pid=$(pidof R)
$ echo $pid
26323

$ cat /proc/26323/cmdline
/usr/lib/R/bin/exec/R--args-emessage('hello there')

Unix is neat

/Henrik
Rolf Turner
#
On 12/01/17 16:33, Henrik Bengtsson wrote:
<SNIP>
Indeed.  Couldn't agree more.  Thanks for the insight.

<SNIP>

cheers,

Rolf
1 day later
Benjamin Tyner
# 
Thank you for the insights, Rolf and Henrik.

To give another example, this time in non-interactive mode,

    Rscript -e "file.exists(commandArgs(TRUE))" <(echo "Hi")

    [1] TRUE

versus

    Rscript -e "normalizePath(commandArgs(TRUE))" <(echo "Hi")
    [1] "/dev/fd/63"
    Warning message:
    In normalizePath(commandArgs(TRUE)) :
      path[1]="/dev/fd/63": No such file or directory

It almost seems like file.exists and normalizePath use separate criteria 
for determining existence?

Regards

Ben
On 01/12/2017 01:42 AM, Rolf Turner wrote:
Jeff Newmiller
# 
Of course they have separate criteria for determining existence... they do different things, and therefore have different requirements for access permissions.