- Package:
- src:edac-utils
- Source:
- edac-utils
- Submitter:
- Martin Pitt
- Date:
- 2021-08-10 22:03:02 UTC
- Severity:
- important
Package: edac-utils Version: 0.16-1 User: mpitt@debian.org Usertags: libsysfs-deprecation Hello, Some years ago libsysfs (source package: sysfsutils) was written as an abstraction layer for accessing /sys/. However, this turned out to be a historical error and evolutionary dead end: It does not actually abstract anything (it's just as specific to the Linux kernel and a particular version thereof as /sys itself), and just adds unnecessary complexity, RAM overhead, and bugs. Thus its development has ceased years ago, in favor of programs just using /sys as it is. In fact, most applications probably don't want to access /sys at all, but use libudev [1] or gudev [2] instead. These provide a better API for device enumeration, properties, and callbacks for hardware changes. This package is one of the few which still use the old libsysfs. Can you please check with upstream to prepare a migration away from libsysfs to using plain /sys or libudev? I hope that we can drop the old libsysfs entirely for wheezy. Thank you for considering! Martin [1] http://www.kernel.org/pub/linux/utils/kernel/hotplug/libudev/ [2] http://www.kernel.org/pub/linux/utils/kernel/hotplug/gudev/
Yaroslav Halchenko <sf@onerussian.com> writes: It is on the TODO list to remove the dependency on libsysfs going forward. It is just a matter of time. Edac-utils has gotten very little development time over the past few years. Going forward there is a push to rearchitect edac core[1] and so I think when that makes it upstream, there should be a complete rewrite of edac-utils or a replacement of something providing similar functionality (command line tool and library for monitoring tools) So does it still make sense to do the work of removing libsysfs from edac-utils, or should we work on a new utility? mark [1] See e.g. https://lkml.org/lkml/2012/5/18/263
Hi! I have to disagree with the above. It abstracts the access to /sys in a more natural form than directly having to do so. Development is also still going, and upstream has recently released a new version merging all Debian patches plus other cleanup and fixes sent by me and others. If this project would be better suited with some other library, then sure go ahead, but please take into account that as the current sysfsutils maintainer in Debian, I consider the library supported and definitely not obsolete, so if you'd like you keep using it, please feel free to do so (and to close this report :). I have no plans for this. See above, if you think another library would suit the project better, then sure, otherwise please feel free to keep relying on the libsysfs library. Thanks, Guillem