bug-xorriso
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[Bug-xorriso] Crash in xorriso with fix


From: Thomas Schmitt
Subject: [Bug-xorriso] Crash in xorriso with fix
Date: Wed, 07 Aug 2019 11:57:18 +0200

Hi,

(It might be that i messed up the moderation of your message.
 I accepted it and cleared the "moderation" flag for your account.
 But now the archive has an empty message from "Unknown".
 If your message does not show up in the next hours then i will
 repost it. For now i add you to Cc: although you are subscribed.)

Pavel Cahyna wrote:
> We were actually interested mainly about the memory management problem
> and not so much about the actual functionality of supporting such huge
> ACL lists, that's why our test was minimal and not realistic,

It served its purpose.
The corrections were necessary only to explore the further stages of this
obviously never heavily challenged code path.


> I am actually surprised you did not hit some limit in ISO9660.

That's because xattr and ACL on ISO 9660 is entirely my own invention.
See
  https://dev.lovelyhq.com/libburnia/libisofs/raw/master/doc/susp_aaip_2_0.txt

libisofs is currently the only software which can encode and decode
them. I.e. you need to use xorriso commands like -extract while -acl is
set to "on" in order to restore ACL from ISO to your hard disk filesystem.
In general ACL are not very portable between different computers. So this
feature is mainly of interest with creating and restoring backups on the
same machine or its appropriately set-up successor.

The implementation can stand giant ACLs, but becomes very slow because
it was originally planned to be only a proof of concept. Manipulation
of xattr and ACL during ISO creation could need a more sophisticated
approach which is faster with finding and inserting data items.

In practice xattr and ACL of files are rather small. So the main waste
of CPU cycles by libisofs is with O(n^2) of its directory insertion
algorithm. But what are 15 seconds of tree building time when afterwards
the BD-RE medium needs 40 minutes to take the ISO image ?
So i wait for my motivation to increase by some real world example that
needs improvements of the suboptimal implementations.


Have a nice day :)

Thomas




reply via email to

[Prev in Thread] Current Thread [Next in Thread]