Hacktoberfest 2026 : les issues que les mainteneurs ont marquées pour octobre, ouvertes et accessibles aux débutants. Parcourir les issues Hacktoberfest

link/unlink is exception unsafe

Ouverte
#286 10 commentaires 0 réactions 1 personne assignée Voir sur GitHub

@hyperthunk y travaille déjà.

Depuis le 9/11/2018.

Évaluation

Cette issue n'a pas encore été évaluée.

Description

task

Suppose I have the following code

f pid = do
  link pid
  blocking_operations_which_depend_on_health_of pid
  unlink pid

If f is interrupted with an exception, unlink might never be called. Thus I rewrite to:

f pid = bracket_ (link pid) (unlink pid) $
          blocking_operations_which_depend_on_health_of pid

But this still doesn't help, because unlink is interruptible, thus if unlink is interrupted with an exception, the process will remain linked to pid after leaving f.

How about

f pid = bracket_ (link pid) (uninterruptibleMask_ $ unlink pid) $
          blocking_operations_which_depend_on_health_of pid

?

This introduces the potential for blocking indefinitely. If pid dies while unlink is executing, the link exception cannot be thrown because uninterruptibleMask_ is in effect. However, when unlink completes, the exception cannot be thrown either, because the very unlink definition does not allow it. The only way to avoid illegal behavior would be to have unlink never return or somehow don't throw the link exception (the latter looks impossible to me).

The second attempt is the one I would expect to work. The problem is determining what tweak of unlink semantics would make it to work.

I would like unlink to be uninterruptible, but I don't see how that could be implemented. An alternative is to guarantee that when unlink runs with exceptions masked, it always unlinks before returning, but it might return with an exception if it is interrupted during its execution.

Langage dominant
Haskell
Étoiles
754
Forks
101
Merge moyen
1 j 3 h
PR mergées (30 j)
4

Préparer son environnement

Par où commencer

  1. Lisez l'issue en entier, puis le guide de contribution du projet.
  2. Signalez en commentaire que vous la prenez — cela évite que deux personnes fassent le même travail.
  3. Forkez le dépôt et travaillez sur une branche.
  4. Ouvrez une pull request qui référence le numéro de l'issue.

Autres issues de haskell-distributed/distributed-process

Toutes les issues de haskell-distributed/distributed-process

Issues similaires

Plus d'issues Haskell

Recevez les nouvelles issues par e-mail

Un résumé court des issues GitHub adaptées aux débutants.