Codi does not output anything and logs: The file "/tmp/cmd" is not executable by this user about codi.vim HOT 19 CLOSED

But internally there's no reason for codi.vim to use fish (or whatever the default 'shell' is). It can just call /bin/bash or env bash or whatever, with jobstart(['env', 'bash', ...]). (Doesn't require modifying user's 'shell'.)

from codi.vim.

This is almost definitely caused by #61. Can you see if reverting 25d4742 fixes it on your end?

from codi.vim.

Well, shellescape shouldn't be used with writefile(). writefile() isn't using a shell...

from codi.vim.

Could be related to fish. Similar problem was reported here: justinmk/vim-dirvish#73

@decent1 If you change your shell to bash (including 'shellcmdflag' and all other shell-related options), does it help?

from codi.vim.

@justinmk, @cympfh had an issue where their version of script couldn't find the python command ($PATH issues maybe?).

EDIT: I see what you mean, oops. I'll try that out.

from codi.vim.

Actually, shellescape() should still probably used since the arguments to the interpreter (python, etc.) still need to be "shell-escaped."

from codi.vim.

(Should add... a lot of these problems are caused by relying on the external script executable, which differs in implementation across platforms. Have been planning a rewrite using Python's pty capabilities which should eliminate platform differences, but school's got me busy.)

from codi.vim.

Ok, so first I tried reverting the 25d4742 commit and that didn't help.

While leaving the commit reverted, I changed my default shell back to bash with chsh -s /bin/bash and relogged (:echo $SHELL outputted /bin/bash), but that didn't help either. Log said the same thing:

[10:25:20.554876] <SNR>131_codi_handle_data[9]: [DATA] bash: /tmp/cmd: Permission denied

Then I tried chmod 777 /tmp/cmd and Codi started working again! For some reason it doesn't work anymore with fish. (I'll soon try again with fish as I really want to use it instead of bash). I hard resetted the revert and as it didn't seem to have an effect one way or the other.

But is that permission denied thing a bug? I tried deleting /tmp/cmd and Codi just recreated it with -rw-rw-r-- permissions, which gives the permission denied.

EDIT: Tried again with fish, with and without the 25d4742, but no luck. I also updated my vim to 8.0.339 and that didn't help either.

from codi.vim.

Sorry I haven't commented on this issue in awhile. @decent1, would you mind pinning down the exact commit in which things broke?

from codi.vim.

@metakirby5 it seems to be commit 9a570ec. If I reset to one commit before it, Codi works again.

from codi.vim.

Ah ok, so it was that commit. I got the chronological ordering of the commits wrong so I thought that reverting 25d4742 would also revert 9a570ec. Mind posting your script version info?

from codi.vim.

script from util-linux 2.27.1

from codi.vim.

Same issue here. Content of /tmp/cmd is 'python', and from the log:

[19:48:28.281643] <SNR>50_codi_handle_data[9]: [DATA] zsh:1: permission denied: /tmp/cmd

After doing chmod u+x /tmp/cmd codi starts responding. Not sure why it needs the file to be executable though.

from codi.vim.

My solution was to create the function below:

function s:AllowCodi()
   if empty(glob('/tmp/cmd'))
       call system('touch /tmp/cmd')
   endif

   call system('chmod u+x /tmp/cmd')
endfunction

And then I called it on VimEnter:

autocmd VimEnter * call s:AllowCodi()

from codi.vim.

Had the same problem, this solution fixed it but should probably be included into the plugin.

Here is with Fedora 26 with script from util-linux 2.30

from codi.vim.

Sorry for the delay. I've added a setfperm call at 93cfc91. Please check out the cmd-perm branch and see if it fixes the issue. I can't test Linux at the moment, but I would like to confirm that it works before merging it.

from codi.vim.

Tried with cmd-perm branch and I don't get the permission error now. But I still don't get any output, at least with Fish shell. The log just says:

[09:49:29.840972] <SNR>154_user_au[1]: Doing autocommand CodiUpdatePre
[09:49:29.841350] <SNR>154_codi_do_update[29]: Starting job for buffer 1
[09:49:29.843900] <SNR>154_stop_job_for_buf[8]: Stopping job for buffer 1
[09:49:29.844653] <SNR>154_codi_do_update[65]: [INPUT] 5 + 5^@^M^D^D^M
[09:49:29.845605] <SNR>154_codi_do_update[66]: Expecting 2 prompts
[09:49:29.852668] <SNR>154_codi_handle_data[9]: [DATA] 5 + 5
[09:49:29.853246] <SNR>154_codi_handle_data[9]: [DATA] Session terminated.

But, if I checkout to commit e2aa007, Codi starts working again.

from codi.vim.

@decent1 Sorry, I had forgotten about the fish issue. What happens if you run /tmp/cmd directly from your shell?

from codi.vim.

~ » /tmp/cmd
Failed to execute process '/tmp/cmd'. Reason:
exec: Exec format error
The file '/tmp/cmd' is marked as an executable but could not be run by the operating system.

I'm noob with linux, so I don't know if this helps, but fish says this when the beginning of the file is missing shebang and it won't know what to do with the contents.

from codi.vim.