Process substition in BASH fails.

[kherge]

Example Script

#!/usr/bin/env bash

while read LINE; do
    echo “$LINE”
done < <(echo hello)

Expected Result

$ hello

Actual Result

line 5: /dev/fd/62: No such file or directory

References

1. http://www.gnu.org/software/bash/manual/html_node/Process-Substitution.html#Process-Substitution
2. https://stackoverflow.com/a/2443120

[tbodt]

#148

[tbodt]

What's missing now is the very weird (to me) linux behavior where pipe file descriptors appear to be broken symlinks but can actually be opened, and become a second file descriptor for the same pipe.

[xinity]

@tbodt any hints to fix this ? as an example we can't leverage powerline10k or any advanced (and very useful features) of ZSH :(

Do you need more tests from us ?
let me know :)

Keep up the awesome work, discovered your app few days ago , NIFTY !!!!

[tbodt]

The problem is well understood (#164 (comment)) but I am not quite sure how to solve it.

[ghost]

This is pretty much over my head but I've been looking around a bit and I stumble upon microsoft/WSL#266 which addresses the same thing(?). I'm not sure if it's helpful, but maybe...

[tbodt]

Yeah, WSL had the same problem. Not much info there about how they solved it though.

[kherge]

I apologize that this is also over my head, but maybe it's buried in one or more commits for their kernel? I tried to create a narrow search that could hopefully be a starting point.

[tbodt]

The bug report linked is for WSL 1, which is not open source. This was never a problem in WSL 2.

[ghost]

I wish I was a developer... For what it's worth I asked in the WSL issue if there would be any possibility for them to share how it was solved. I'm not sure how I'm able to help in any other way. The worst that can happen is no one answers or we get a no.

[makakin]

Any solution to fix this issue?

[le0-VV]

Is this thread still going? Came across this problem and is now mildly infuriating me

[luyui]

What's missing now is the very weird (to me) linux behavior where pipe file descriptors appear to be broken symlinks but can actually be opened, and become a second file descriptor for the same pipe.

@tbodt just ran across this issue and dropped a dirty quick fix at #1716
The basic idea is to reuse adhoc_fd and avoid procfs handling the io, thus done some trick in struct mount *find_mount_and_trim_path(char *path) .
There might be rough edges though, but at least these substitution test cases now run without error.

[ryleu]

@tbodt there seems to be a solution now, so I'd like to bring this to your attention again.

[annpocoyo]

What’s weird is that the pipe still gets created.
As seen by running ls /proc/self/fd <(echo “test”), even if you make a symlink to /proc/self/fd from /dev/fd as seen by running ls /dev/fd <(echo “test”). Someone please merge the pull request

[kalebmckale]

Is this related? https://gitlab.alpinelinux.org/alpine/aports/-/issues/1465

[kalebmckale]

@annpocoyo: I'm seeing this on mine:

  File: '/dev/fd/63' -> 'anon_inode:[unknown]'
  Size: 0               Blocks: 0          IO Block: 0      symbolic link
Device: 0h/0d   Inode: 17732923551252482  Links: 0
Access: (0777/lrwxrwxrwx)  Uid: (    0/    root)   Gid: (    0/    root)
Access: 1970-01-01 00:00:00.000000000 +0000
Modify: 1970-01-01 00:00:00.000000000 +0000
Change: 1970-01-01 00:00:00.000000000 +0000

when I run stat. The symlink is indeed created, but it's not linked to a pipe as it should be (from what I gather from a bunch of reading I've been doing).

More so, I noticed strangely that running processes in general are not being created as one would expect:

> sleep 1000 &
[1] 221
> lsof -p 221
COMMAND PID USER   FD      TYPE DEVICE   SIZE    NODE NAME
sleep   221 root  cwd   unknown                       /proc/221/cwd (readlink: No such file or directory)
sleep   221 root  rtd   unknown                       /proc/221/root (readlink: No such file or directory)
sleep   221 root  txt       REG    1,8 808104 4317863 /bin/busybox
sleep   221 root    0u      CHR  136,0              3 /dev/pts/0
sleep   221 root    1u      CHR  136,0              3 /dev/pts/0
sleep   221 root    2u      CHR  136,0              3 /dev/pts/0

I'm not sure this is exactly related but since I discovered this during my reading and research, I figured it might be useful?

[rrr9rrr]

My testing results:

iPhone7nABEL:~# ls -la <(:) /proc/self/fd
lrwxrwxrwx 0 root root 0 Jan 1 1970 /dev/fd/63 -> 'anon_inode:[unknown]'

/proc/self/fd:
total 0
lrwxrwxrwx 0 root root 0 Jan 1 1970 0 -> /dev/pts/0
lrwxrwxrwx 0 root root 0 Jan 1 1970 1 -> /dev/pts/0
lrwxrwxrwx 0 root root 0 Jan 1 1970 2 -> /dev/pts/0
lrwxrwxrwx 0 root root 0 Jan 1 1970 3 -> /proc/35/fd
lrwxrwxrwx 0 root root 0 Jan 1 1970 63 -> 'anon_inode:[unknown]'
iPhone7nABEL:# ls -la <(:) /dev/fd
lrwxrwxrwx 1 root root 13 Nov 29 07:41 /dev/fd -> /proc/self/fd
lrwxrwxrwx 0 root root 0 Jan 1 1970 /dev/fd/63 -> 'anon_inode:[unknown]'
iPhone7nABEL:#

iSH 1.3.2 (Build 494)