RPM

race in filetriggers

Bug #911007 reported by Jeff Johnson
6
This bug affects 1 person
Affects Status Importance Assigned to Milestone
RPM
New
Undecided
Unassigned
ALT Linux
Unknown
Unknown

Bug Description

Running two apt-get instances in parallel results in
...
1: kernel-image-std-def
########################################################################
[100%]
/
Running /usr/lib/rpm/posttrans-filetriggers
warning: failed to restore database lock

(with 'apt-get install update-kernel' running in parallel)

After this apt-get says that kernel-image is installed, while rpm says that it is not.
Files from the package do exist in the system

This problem can be likely solved by cleaning apt caches, though this is not convenient.

Tags: triggers alt
Jeff Johnson (n3npq)
tags: added: alt triggers
Denis Silakov (dsilakov)
description: updated
summary: - race в filetriggers
+ race in filetriggers
Revision history for this message
Jeff Johnson (n3npq) wrote :

Thank you for the translation.

There are 2 coupled issues that are pertinent to ROSA planning here
(in spite of this being an Alt bug on a heavily patched older version of RPM)

1) file pattern triggers
     rpm5.org is carrying 2 file trigger implementations at the moment.

     One of the implementations originally came from Mandriva (and might
     be what is back ported to Alt, I haven't looked).

     There's another implementation (from me) that extends %trigger syntax
     to use patterns on file/directory paths. The %trigger implementation
     was undertaken so that the scripts that run are directly present in
     *.rpm packaging and fit into existing *.spec syntax in a "natural" way. The
    Mandriva file triggers are entirely opaque to rpm which merely reads
    patterns and runs scripts. The problem is that rpm gets blamed for
    configuration failures because of the mechanism and its rather difficult
    QA to find/fix reproducers when they occur.

     One of Per Oyvind's wiki page bugs is related to this same issue.

2) apt-rpm support

     Yes, the "cache" in apt leads to a class of problems that are similar
     to the rpmdb "database" problems routinely reported as rpm "bugs".

     Meanwhile -- if ROSA is considering distributing/using/supporting apt-rpm,
     then blueprints and planning needs to be undertaken.

To post a comment you must log in.
This report contains Public information  
Everyone can see this information.

Other bug subscribers

Remote bug watches

Bug watches keep track of this bug in other bug trackers.