The long awaited s3cmd 0.9.8.4 has finally been released with the following enhancements:
- Re-issue temporarily failed requests – This should get rid of all those annoying “Connection resets”, “Server internal errors”, etc, occasional temporary errors. Should a request fail it will be re-issued up to 5x.
- Restored access to upper-case named buckets – However Amazon doesn’t allow upper-case buckets in EU datacentre.
- Improved handling of filenames with non-ascii characters. – This is a long lasting story and every time there is someone who complains that their national alphabet is not handled correctly by s3cmd. Since 0.9.8.4 all filenames are (should be) internally converted to Unicode. Hope that will sort most problems out. This change should have no effect if you only use basic US alphabet.
- Doesn’t run out of open files quota on sync with lots of files. – When syncing lots of files from S3 to local directory some users ran out of file handles. This shouldn’t happen anymore.
- Sync skips over files that can’t be open instead of terminating the sync completely. – Self explanatory.
- Avoid ZeroDivisionError on ultrafast links (for instance on Amazon EC2) – Also self explanatory I believe.
- A couple of small bug fixes
In other words s3cmd 0.9.8.4 should be much more stable and resilient to temporary errors then any earlier version. All users of previous versions should upgrade.
Thanks to everyone who provided feedback, bug reports and patches. I appreciate your effort to make s3cmd better!
Source package is available on SourceForge.net downloads page
Alternatively check out the other ways to get the sources on our Downloads page.
If you’re using openSUSE, SUSE Enterprise, Fedora, RedHat Enterprise or CentOS the best way to keep your s3cmd up to date is to add s3cmd repository to your ‘yum’ or ‘yast’ config. Check out our page about Package Repositories for more details.
Please report any issues encountered to s3tools-general@lists.sourceforge.net mailing list.
Enjoy!
Michal Ludvig
