We've published a new maintenance release 14.2.5 in our subscription repositories, the source code is as always available on GitHub.
This release contains several bugfixes, see the bareos-14.2.5 release ticket.
Summary of fixes:
- Do not assume a filetype in digest_file
- Unfortunately, the ff_pkt->type is not set correctly when
digest_file is called during accurate comparison.
We get the value that stored before in ff_pkt->type by the last
file/directory that was processed, not by what we actually see.
That is why we do always calculate the digest now.
#462: windows file daemon and zero-length files in Inc backup Description
- Treat deduplicated reparse points correctly
- IO_REPARSE_TAG_DEDUP reparse points were treated like symlinks before.
We now understand these reparese points and treat them like normal
files, which is called
"Nonoptimized Backup and Restore of Data Deduplication-Enabled Volumes"
This means the files are treated like normal files and not like reparse
points.
Files on deduplicated volumes can be backed up and restored like files
on any other windows filesystem.
For more details see:
https://msdn.microsoft.com/en-us/library/hh769304%28v=vs.85%29.aspx
#467: Error in restoring deduplicated files
- Dir crashes if media.ScratchPoolId = NULL
- Wrong array indices while checking for NULL corrected.
#469: Dir crashes if media.ScratchPoolId = NULL
- Fix config pretty printer.
- When printing the compression used it eats an extra char so the pretty
printer doesn't print the next option in the option string. We should
also print GZIP without \n as we print the level after it, currently
its printed on the next line.
#466: option noatime = yes in bconcole displays like 6 (digit)
- Fix bareos logrotate errors
- #417: bareos logrotate errors
- Fix restore of compressed files
- The compressed file restoration failed before on Windows 2008.
We now use the variable types from the microsoft documentation,
also now have a bytesreturned variable which is mandatory
if lpOverlapped is NULL according to the documentation.
See documentation of FSCTL_SET_COMPRESSION control code at Microsoft:
https://msdn.microsoft.com/en-us/library/windows/desktop/aa364592%28v=vs.85%29.aspx
#463: Crash when restore compressed files
- Rerun overhaul
- - Only allow rerun of Job types that make sense e.g.
- Backup
- Copy
- Migration
- Rerun now also works with virtual full backups.
We need to set the pool to the original sourcepool,
not to the pool that is saved in the job record
as that is the nextpool.
Therefore we get the pool from the job definition
so that the virtual full backup can be rerun
- Only set successful Jobs to L_FULL when running as L_VIRTUAL_FULL
This saves us from doing all kind of black magic along the way.
Set jr.JobLevel to L_FULL only for successful jobs.
#370: rerun since_jobid= runs Virtual Fulls as Fulls.
- Always use new block for autoexclude
- #415: For Windows clients not working include, only exlude works
- Fix accurate mode when using strip path.
- #366: Accurate Mode and strippath
- Change the order of output for list jobs
- We now ORDER BY JobId instead of
ORDER BY StartTime,JobId
This prevents showing jobs without starttime timestamp
always as last output.
#403: "list jobs" command output is not always sorted
- Use proper variable type for format attribute.
- DeviceIoControl with flag FSCTL_SET_COMPRESSION expect the format to be
as unsigned short not integer which can be 32 or 64 bits depending on
the compile. Switched this variable to uint16_t which should be an
unsigned short.
#436: Restore crash file daemon
- Don't crash when ua->UA_sock == NULL in prtit()
- Some admin jobs have an UA context but not a ua->UA_sock and when you
then blind use ua->UA_sock->fsend() you crash in the worst possible way.
We now call ua->send_msg() which has some fallback logic when
ua->UA_sock is NULL and then uses Jmsg with M_INFO to redirect the info
to the Job.
Its might not fully fix #418 as its seems the admin Job want to have
interaction with the user which ain't going to work in an admin Job but
crashing is about the worse what can happen.
#418: traceback runnning after job
- Allow TLS 1.2 connections when using openssl
- Adding TLS 1.2 support will allow us to be more future-proof and have
better ciphersuites such as as the use of ECDHE-ECDSA-AES256-GCM-SHA384.
This patch allows tls 1.2. 1.1 and 1.0 while the broken sslv2 and
sslv3 are disabled.
#440: Allow TLS 1.2 connections when using openssl
- Add escape parsing to generic plugin option parser.
- Backport the new escaping of the ':' using '\:' to all existing plugin
config parsers. This code comes from the new cephfs-fd and gfapi-fd
plugins.
#428: bpipe plugin on windows and colon character problem
- Fix use of wrong variable for WeekOfMonth check in next hour.
- wom ==> nh_wom
#424: Scheduling select wrong run line in schedule.
Please refer to the documentation: