Ticket #116 (new defect) — at Initial Version

Opened 15 years ago

Last modified 10 years ago

savannah: infinite loop reading large directories via fish

Reported by: slavazanko Owned by:
Priority: major Milestone: Future Releases
Component: mc-vfs Version: master
Keywords: Cc: zaytsev
Blocked By: Blocking:
Branch state: no branch Votes for changeset:

Description

Original: http://savannah.gnu.org/bugs/?15801

Submitted by:Mario Lorenz <mlo>Submitted on:Sun 19 Feb 2006 12:15:18 PM UTC
Category:VFSSeverity:3 - Normal
Status:In ProgressPrivacy:Public
Assigned to:Pavel Tsekov <ptsekov>Open/Closed:Open
Release:4.6.1Operating System:GNU/Linux

Discussion:

Thu 23 Feb 2006 03:38:12 PM UTC, comment #1:

This problem has been bugging me for a while. I've just commited a patch which exposes a new user configurable option:

fish_directory_timeout

It contains the lifetime of a directory cache entry measured in seconds. I've adjusted the default value to 900 seconds (same as in ftpfs).

This option is not configurable through the user interface, yet - one can change it only by directly editing MC's ini file. I plan to fix this soon.

To test the new code you need to fetch MC from the cvs repository or grab a snapshot.
	Pavel Tsekov <ptsekov>
Project AdministratorIn charge of this item.
Sun 19 Feb 2006 12:15:18 PM UTC, original submission:

Reading large remote directories via fish (shell link)
over slow network links causes an infinite or at least very long
loop when mc tries to read the directory multiple times.

This is due to the fish directory timeout being hardcoded to 10
seconds, whereas reading a 15000 entry directory via a 64kbit/s link will take two minutes (way longer if not using compression). This means the directory objects will be marked obsolete
before the directory is even loaded, causing an immediate reload once finished, with this pattern sometimes repeating even more often.

That timeout should be tied to the (user settable) ftp directory timeout, or be given its own user settable value; at the very least it should be set to a sane value (that is, >> 10 seconds) 
Note: See TracTickets for help on using tickets.