Got "gitlab.nvim: Could not run command!" when start review

[sergey-arkhipov]

Prerequsities

• I'm on the latest version of the plugin
• I've installed the required dependencies
• I've run :h gitlab.nvim.troubleshooting and followed the steps there

Setup Configuration and Environment

return {
  'harrisoncramer/gitlab.nvim',
  dependencies = {
    'MunifTanjim/nui.nvim',
    'nvim-lua/plenary.nvim',
    'sindrets/diffview.nvim',
    'stevearc/dressing.nvim', -- Recommended but not required. Better UI for pickers.
    'nvim-tree/nvim-web-devicons', -- Recommended but not required. Icons in discussion tree.
  },
  enabled = true,
  build = function()
    require('gitlab.server').build(true)
  end, -- Builds the Go binary
  config = function()
    require('gitlab').setup({
      discussion_tree = { -- The discussion tree that holds all comments
        auto_open = false, -- Automatically open when the reviewer is opened
      },
    })
  end,
}

Bug Description

Started got described error message

Reproduction Steps

1. Simply start review or another action, edit summary, for example

Screenshots

[sergey-arkhipov]

Additional info

The problem start after install last upgrade for

• 'MunifTanjim/nui.nvim'
• 'nvim-lua/plenary.nvim'

Rollback to previous version (commit) fix problem

    { 'MunifTanjim/nui.nvim', commit = '9e3916e' },
    { 'nvim-lua/plenary.nvim', commit = '2d9b06177a975543726ce5c73fca176cedbffe9d' },
....

[jakubbortlik]

Hi @sergey-arkhipov, can you confirm that this can actually be narrowed down to the update in plenary.nvim? I don't think nui.nvim has to be pinned to 9e3916e to get rid of the issue.

I also don't think this is a bug in gitlab.nvim. If I increase the timeout in the Job that is causing Could not run command! like this:

diff --git a/lua/gitlab/job.lua b/lua/gitlab/job.lua
index 7f5f4d8..cff79d3 100644
--- a/lua/gitlab/job.lua
+++ b/lua/gitlab/job.lua
@@ -53,7 +53,7 @@ M.run_job = function(endpoint, method, body, callback)
           local message = string.format("%s: %s", data.message, data.details)
           u.notify(message, vim.log.levels.ERROR)
         end
-      end, 0)
+      end, 10)
     end,
     on_stderr = function()
       vim.defer_fn(function()

then the review is started normally, which suggests the on_stdout (lua/gitlab/job.lua line 23) is executed, but I'm still getting Could not run command! which means the on_stderr callback (lua/gitlab/job.lua line 58) is executed as well.

[sergey-arkhipov]

My conclusion is based on the following facts

1. gitlab.nvim works fine in my configuration
2. Perform a regular update neovim packages
3. Get the error described.
4. Look at the dependencies that were updated and could affect the behavior.
5. Specify previous versions of dependencies in the gitlab.nvim configuration.
6. Get a working structure without errors.
   That's all. :)

[jakubbortlik]

Sure, that's a perfectly valid process. Can you please help debug this by confirming that the issue is not really with nui.nvim?

[sergey-arkhipov]

Can you please help debug this by confirming that the issue is not really with nui.nvim?

I would be happy if you could describe the process step by step, what needs to be done.

[jakubbortlik]

Just use { 'MunifTanjim/nui.nvim'}, instead of { 'MunifTanjim/nui.nvim', commit = '9e3916e' }, and only pin plenary.nvim to the working version.

[sergey-arkhipov]

Yes, you are right. This also works without errors.

[jakubbortlik]

The issue can be fixed with this patch:

diff --git a/lua/gitlab/job.lua b/lua/gitlab/job.lua
index 7f5f4d8..d93730e 100644
--- a/lua/gitlab/job.lua
+++ b/lua/gitlab/job.lua
@@ -55,10 +55,12 @@ M.run_job = function(endpoint, method, body, callback)
         end
       end, 0)
     end,
-    on_stderr = function()
-      vim.defer_fn(function()
-        u.notify("Could not run command!", vim.log.levels.ERROR)
-      end, 0)
+    on_stderr = function(err, data)
+      if err or data then
+        vim.defer_fn(function()
+          u.notify("Could not run command!", vim.log.levels.ERROR)
+        end, 0)
+      end
     end,
     on_exit = function(_, status)
       vim.defer_fn(function()

@harrisoncramer, I believe it would make sense to actually print the error that caused the command not to run, what do you think?

[sergey-arkhipov]

The issue can be fixed with this patch:

Thank you very much. I checked the attached patch, everything works correctly.
I can make a PR for changes if necessary.

return {
  'sergey-arkhipov/gitlab.nvim',
  dependencies = {
    { 'MunifTanjim/nui.nvim' },
    { 'nvim-lua/plenary.nvim' },
    'sindrets/diffview.nvim',
    'stevearc/dressing.nvim', -- Recommended but not required. Better UI for pickers.
    'nvim-tree/nvim-web-devicons', -- Recommended but not required. Icons in discussion tree.
  },
...

gitlab.nvim   main  gd HEAD~1
diff --git a/lua/gitlab/job.lua b/lua/gitlab/job.lua
index 7f5f4d8..d93730e 100644
--- a/lua/gitlab/job.lua
+++ b/lua/gitlab/job.lua
@@ -55,10 +55,12 @@ M.run_job = function(endpoint, method, body, callback)
         end
       end, 0)
     end,
-    on_stderr = function()
-      vim.defer_fn(function()
-        u.notify("Could not run command!", vim.log.levels.ERROR)
-      end, 0)
+    on_stderr = function(err, data)
+      if err or data then
+        vim.defer_fn(function()
+          u.notify("Could not run command!", vim.log.levels.ERROR)
+        end, 0)
+      end
     end,
     on_exit = function(_, status)
       vim.defer_fn(function()

[jakubbortlik]

Well, that was just a half-baked solution. In reality something more elaborate would be necessary, something like the following, where the individual lines from stderr are collected and only logged once at the end of the process:

diff --git a/lua/gitlab/job.lua b/lua/gitlab/job.lua
index 7f5f4d8..128591b 100644
--- a/lua/gitlab/job.lua
+++ b/lua/gitlab/job.lua
@@ -17,6 +17,7 @@ M.run_job = function(endpoint, method, body, callback)
   -- This handler will handle all responses from the Go server. Anything with a successful
   -- status will call the callback (if it is supplied for the job). Otherwise, it will print out the
   -- success message or error message and details from the Go server.
+  local stderr = {}
   Job:new({
     command = "curl",
     args = args,
@@ -55,13 +56,20 @@ M.run_job = function(endpoint, method, body, callback)
         end
       end, 0)
     end,
-    on_stderr = function()
-      vim.defer_fn(function()
-        u.notify("Could not run command!", vim.log.levels.ERROR)
-      end, 0)
+    on_stderr = function(_, data)
+      if data then
+        table.insert(stderr, data)
+      end
     end,
-    on_exit = function(_, status)
+    on_exit = function(code, status)
       vim.defer_fn(function()
+        if #stderr ~= 0 then
+          u.notify(
+            string.format("Could not run command `%s %s`! Stderr was:", code.command, table.concat(code.args, " ")),
+            vim.log.levels.ERROR
+          )
+          vim.notify(string.format("%s", table.concat(stderr, "\n")), vim.log.levels.ERROR)
+        end
         if status ~= 0 then
           u.notify(string.format("Go server exited with non-zero code: %d", status), vim.log.levels.ERROR)
         end

I'll open a PR and see what @harrisoncramer thinks about it.

[harrisoncramer]

Thank you both for resolving this, I've been busy with a new job recently and haven't been dedicating much time to this plugin. I've merged in Jakub's fix to develop and will ship that into main now.