uutils / uutils/coreutils

Sort tempfile logic opens up extra files

Open
#6,945 4 comments 0 reactions 0 assignees View on GitHub

Nobody has claimed this yet.

U - sort
Dominant language
Rust
Stars
24.1k
Forks
2k
Avg merge
1d 5h
Merged PRs (30d)
365

Description

As part of investigating this issue, found that we're using extra file descriptors to mange our tempfile logic for batched-sort operations. This is a problem because some of the GNU compatibility tests run with ulimit enforcing strict limits on the number of files that sort can open.

Looking at the open files when starting a batched sort, I see...

$ ls -l /proc/1778302/fd
total 0
lrwx------ 1 karl karl 64 Dec  8 19:43 0 -> /dev/pts/3
l-wx------ 1 karl karl 64 Dec  8 19:43 1 -> /home/karl/Rust/gnu/tests/sort/sort-merge-fdlimit.log
l-wx------ 1 karl karl 64 Dec  8 19:43 2 -> /home/karl/Rust/gnu/gt-sort-merge-fdlimit.sh.A1OJ/err/merge-random-err
lr-x------ 1 karl karl 64 Dec  8 19:43 3 -> 'pipe:[3021894]'
l-wx------ 1 karl karl 64 Dec  8 19:43 4 -> 'pipe:[3021894]'

It's the last two of these that are problematic. These two pipes are created when we set up the Tempfile logic...

    fn init_tmp_dir(&mut self) -> UResult<()> {
<snip>
        let path = self.temp_dir.as_ref().unwrap().path().to_owned();
        let lock = self.lock.clone();
// *** The line below will open the two pipes.
        ctrlc::set_handler(move || {
            // Take the lock so that `next_file_path` returns no new file path,
            // and the program doesn't terminate before the handler has finished
            let _lock = lock.lock().unwrap();
            if let Err(e) = remove_tmp_dir(&path) {
                show_error!("failed to delete temporary directory: {}", e);
            }
            std::process::exit(2)
        })
        .map_err(|e| USimpleError::new(2, format!("failed to set up signal handler: {e}")))
    }

To fix this, I propose to just simplify this logic and just remove the Tempdir altogether. Just put any temporary files directly into /tmp (which is what GNU-sort does) and have the OS do the cleanup automatically rather than create a dedicated folder for them. We can then remove this code above that uses the ctrlc crate.

I don't know the history on this, so maybe there's a good reason to use a separate folder - If so, please let me know!

Contributor guide

Open the contributing guide

First steps

  1. Read the whole issue, then the project's contributing guide.
  2. Comment on the issue to say you are picking it up — it saves two people doing the same work.
  3. Fork the repository and make your change on a branch.
  4. Open a pull request that references the issue number.

Research direction

Start at sort's init_tmp_dir logic and trace how Tempfile, the temporary directory, and the ctrlc handler are used during batched sorts. Run the GNU compatibility tests for sort-merge-fdlimit and inspect open descriptors while sorting. Done means the extra pipes are no longer opened and the fd-limit tests still pass.

Written by the indexing model from the issue text.

Assessment

Tech stack
rust
Domain
cli
Issue type
Bug
Difficulty
4/5
Estimated time
3-5 days
Activity status
Stale
Clarity
Mostly clear
Newbie friendliness
42/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.