2.9.0

This release contains code fixes to eliminate compiler warnings (cum errors) on new(er), stricter, standards-enforcing compilers e.g. found in Debian 9.2 and up or GCC7.*.

New features added: remote mount/unmount of Mark5 disk packs as if the disk pack was keyed on/off.

Finer grained hardware detection+checking to support 'crippled' Mark5's - systems with only a StreamStor card. All disk-based functionality remains working but recording (you need an I/O board or daughter card for that).

Recording data rate check implemented to prevent data corruption.

For full details see the changelog

SDK9.2.1 version: jive5ab-2.9.0-ETCH-SDK921-i386.deb  

64 bit SDK9.4 version: jive5ab-2.9.0-WHEEZY-SDK94-amd64.deb  

Source code: jive5ab-2.9.0.tar.gz

2.8.1

This is a bugfix update for 2.8.0, no new developments. Important fixes include but are not limited to:

There is a multi-thread race condition lurking in FlexBuff "disk2file" and "disk2net" so it is advised to upgrade for non-crashing behaviour.

This version can erase empty disk packs once again (2.8.0 lost this due to overfocus on non-bank mode support)

The "record?" reply on FlexBuff was so broken by 2.8.0 that the Field System correctly complains loudly and bitterly

For full details see the changelog

SDK9.2.1 version: jive5ab-2.8.1-ETCH-SDK921-i386.deb  

64 bit SDK9.4 version: jive5ab-2.8.1-WHEEZY-SDK94-amd64.deb  

64 bit NO STREAMSTOR: jive5ab-2.8.1-WHEEZY-NOSSAPI-amd64.deb  

Source code: jive5ab-2.8.1.tar.gz

2.8.0

This is the first stable release since 2.7.1 (~ a year ago!), having something for everybody:

• high-resolution time stamps, supports at least 256Gbps/bit stream
• use 'net_protocol=udpsnor' to capture multiple FiLa10G VDIF streams
• "scan_set=XXXX" auto-detects FlexBuff or Mark6 recording format
• workable non-bank mode on all Mark5 systems; BUT SEE SECTION 11 in DOCUMENTATION v1.7!! Use e.g. for 4Gbps on Mark5C
• Python utils m5copy, SSErase.py, DirList.py updated to deal with target Mark5 in non-bank mode
• UDT performance significantly enlarged, especially on long, fat links
• transient runtimes: resources automatically cleaned up if control connection closed; m5copy updated to use this feature
• fix two SEGFAULTs (a data race condition and we were triggering undefined behaviour)

SDK8.2 version: jive5ab-2.8.0-ETCH-SDK82-i386.deb  

SDK9.2.1 version: jive5ab-2.8.0-ETCH-SDK921-i386.deb  

64 bit SDK9.4 version: jive5ab-2.8.0-WHEEZY-SDK94-amd64.deb  

64 bit NO STREAMSTOR: jive5ab-2.8.0-WHEEZY-NOSSAPI-amd64.deb  

Source code: jive5ab-2.8.0.tar.gz

2.7.1-ack-udfix

This unplanned release is a backport to the latest stable branch of jive5ab (2.7.1 branch) of a severe 32 vs 64 bit cross-system portability bug that was first fixed in the development head. Testing of the bugfix revealed two other issues potentially affecting operational use which were also fixed and included in this backport.

On 64 bit systems the compiler rounded the size of the disk info meta data structure to a different size then on 32 bit platforms. Those structures are written verbatim to Mark5A/Mark5B user directory layouts, thus producing differently sized user directories depending on which system wrote them.

The bug means that disk packs erased on one platform might potentially not be recognized by software on another platform: the data becomes mostly inaccessible but to software that completely ignores the user directory. This is because the user directory size is the only key to identifying which permutation it is.

To add insult to injury, nobody seems to really use this data at all ...

This update fixes the code such that on all platforms the structures now have a consistent size. A separate structure was added to support the erroneously sized user directories that may have been written since the release of 64 bit jive5ab.

Note that the actual data and user directory contents were never in danger of being lost, it's just that the software on another platform might not be able to decode the user directory layout.

During testing it was found that failure to create a valid user directory (for any reason) would leave the disk pack without user directory at all and without its given VSN. The fixed behaviour is that a failed attempt to erase will leave the disk pack in its current state.

It was also found that a failed "protect = off;" (return code 4 "error during execution") would leave the system in a state where the next command would be executed as if "protect = off;" had succeeded - i.e. that it was still possible to make modifications to the disk pack. The fixed behaviour is to only clear the protected flag if the "protect = off" command actually succeeds.

SDK8.2 version: jive5ab-2.7.1-ack-udfix-ETCH-SDK82-i386.deb  

SDK9.2.1 version: jive5ab-2.7.1-ack-udfix-ETCH-SDK921-i386.deb  

64 bit SDK9.4 version: jive5ab-2.7.1-ack-udfix-WHEEZY-SDK94-amd64.deb  

Source code: jive5ab-2.7.1-ack-udfix.tar.gz

2.7.1

MIT Haystack Mark6 software writes their data files in a subdirectory called "/data/" on each disk. jive5ab's internal Mark6 aliases for disk modules ("1", "2", "3" and "4") have been updated to reflect this such that a recording's data files can be found. Also when indexing the "eMSN" of a module, the extra "/data/" directory is automatically added.

the indexing of a Mark6 d-plane version 2 format recording has been parallelized per disk rather than doing it sequentially, resulting in linear speedup.

shortly after the release of 2.7.0 it was discovered that changes in the "net2file=" implementation had made it vulnerable to a possibility where under certain circumstances the current runtime would not be able to get out of "net2file" state, rendering it useless. The impact on production systems would range from mildly inconvenient (transient runtimes as used by m5copy) to unacceptable where the only way to regain control was to "^C" jive5ab and restart it.

it turned out that all "runtime=" command processing happened outside an error-catching block. The net effect being that in the rare occassion of a "runtime=" command failing, jive5ab would not crash but perform an orderly shutdown, as if the user had pressed "^C". This unexpected and very unpleasant behaviour is now fixed.

SDK8.2 version: jive5ab-2.7.1-ETCH-SDK82-i386.deb  

SDK9.2.1 version: jive5ab-2.7.1-ETCH-SDK921-i386.deb   jive5ab-2.7.1-LENNY-SDK921-i386.deb

SDK9.3a version: jive5ab-2.7.1-LENNY-SDK93a-i386.deb  

64 bit SDK9.4 version: jive5ab-2.7.1-WHEEZY-SDK94-amd64.deb  

Source code: jive5ab-2.7.1.tar.gz

2.7.0

This version expands FlexBuff and Mark6 usability and contains many bugfixes. Not in the least some fallout from releasing the documentation was dealt with.

Headlines for this release. A detailed overview of all fixes/edits is contained in the changelog file in the distribution or on this webpage (see above).

• implement Mark5 commands scan_set, scan_check, dir_info, rtime, disk2net, disk2file for FlexBuff/Mark6
• changes to allow m5copy to resume a previously started and interrupted transfer
• fix SEGV in the UDT library and work around a libc license-to-segv
• many other small improvements and fixes

It is important to note that 2.7.0 still does NOT support the Mark6 command set. Disk module initialization, mounting and unmounting needs to be done outside jive5ab. Still, no attempt at keeping a "user directory" or grouping of disk modules is made.

SDK8.2 version: jive5ab-2.7.0-ETCH-SDK82-i386.deb  

SDK9.2.1 version: jive5ab-2.7.0-ETCH-SDK921-i386.deb   jive5ab-2.7.0-LENNY-SDK921-i386.deb

64 bit SDK9.4 version: jive5ab-2.7.0-WHEEZY-SDK94-amd64.deb  

Source code: jive5ab-2.7.0.tar.gz

2.6.0

This version of jive5ab contains many small fixes and a big new development. Changes affecting production:

• Fixes in jive5ab for succesful file -> file transfers. An underexplored area as it turned out to be. m5copy has been modified accordingly.
• Fixed problem where correct functioning of Mark5C "mode=" and "clock_set=" commands depended on the order in which they were sent.
• A few users reported incorrect total capacity labels when "vsn=..." on disk packs > 16TB. The recording capacity was never compromised, just the text in the label.
• FlexBuff recording does not write to the root file system any more.

As of 2.6.0 jive5ab can record on Mark6. On both FlexBuff and Mark6 there is a choice of recording in FlexBuff mode or in native Mark6 mode - version 2 of the file format written by the MIT Haystack dplane software.

It is important to note that 2.6.0 does NOT support the Mark6 command set. Disk module initialization, mounting and unmounting needs to be done outside jive5ab. Also no attempt at keeping a "user directory" or grouping of disk modules is made.

SDK9.2.1 version: jive5ab-2.6.0-ETCH-SDK921-i386.deb   jive5ab-2.6.0-LENNY-SDK921-i386.deb

64 bit SDK9.4 version: jive5ab-2.6.0-WHEEZY-SDK94-amd64.deb  

Source code: jive5ab-2.6.0.tar.gz

2.5.1

Significant improvements in specific areas. If it's not listed below, it's not changed. The source code distribution contains the latest version of m5copy (1.17).

• Scan label policy is now enforced. According to Haystack documentation, scan labels should always be EXP_STATION_SCAN formatted. In case this isn't the case, the label will be made to look like that by adding the strings 'EXP_' and 'ST_' as needed.
• Mark5C control from the FS is now possible. Fixed disk2file with time-stamp based data extraction even on (broken) Mark5B formatted Digital Backend output.
• Implemented measures to prevent data loss due to streamstor record pointer being reset.
• Implemented protection against non eight-byte aligned recordings on streamstor AMAZON systems.
• (feature request) print out of HDD properties of detected disks when a disk pack is mounted.
• The 'errors?' (and 'status?' query error bit) will contain useful information if e.g. a background operation fails for some reason. This typically applies to commands that return '1': "action initiated but not completed".
• (initial) support for VDIF detection in "scan_check?". This is done using heuristics because VDIF lacks a syncword which can be used to unequivocally detect the frames without prior knowledge of the VDIF stream
• Improvements to operationalize FlexBuff recordings.
• Improvements to support proper m5copy operation under all use cases
• m5copy gained the ability to read data from the telescope input directly ("in://" source URI) or from the jive5ab internal circular memory buffer, when doing buffered recording ("mem://" source URI).

SDK8.2 version: jive5ab-2.5.1-ETCH-SDK82-i386.deb

SDK9.2 version: jive5ab-2.5.1-ETCH-SDK921-i386.deb   jive5ab-2.5.1-LENNY-SDK921-i386.deb

Source code: jive5ab-2.5.1.tar.gz

2.5.0

Minor update from 2.5.0-beta to official release

• Confusing and erroneous warning message when running on Mark5C fixed
• CONT14 update
  • "reset=erase" defaults to large (64k scans) user directory
  • Support for "reset = erase : s" to force small, 1k scans, user directories

SDK8.2 version: jive5ab-2.5.0-ETCH-SDK82-i386.deb

SDK9.2 version: jive5ab-2.5.0-ETCH-SDK92-i386.deb   jive5ab-2.5.0-LENNY-SDK92-i386.deb

Source code: jive5ab-2.5.0.tar.gz

2.5.0-beta

Features in the 2.5 series of jive5ab include

• Mark5B(+) more descriptive DOT clock errors/status, even stricter error detection and even more predictable "dot_set=" behaviour
• Mark5C support
• Fix a bug in the "get_stats?" query for disk packs not containing 8 disks
• Protection of data transfers against sending arbitrary commands - each command has specific conditions under which it is allowed to execute. We think we have imposed sensible conditions but please report anything that either breaks or does not allow you to do what you need to do
• Modifications to implement all features of the m5copy e-shipping program
• The source code distribution contains the latest Python based utilities SSErase.py, DirList.py and m5copy. See Notes of the 2.4.5 release below for potential additional Python module to install

SDK8.1 version: jive5ab-2.5.0-beta-ETCH-SDK81-i386.deb

SDK8.2 version: jive5ab-2.5.0-beta-ETCH-SDK82-i386.deb

SDK9.2 version: jive5ab-2.5.0-beta-ETCH-SDK92-i386.deb   jive5ab-2.5.0-beta-LENNY-SDK92-i386.deb

Source code: jive5ab-2.5.0-beta.tar.gz

2.4.6

CONT14 update:

• "reset=erase" defaults to large (64k scans) user directory
• Support for "reset = erase : s" to force small, 1k scans, user directories

SDK8.2 version: jive5ab-2.4.6-ETCH-SDK82-i386.deb

SDK9.2 version: jive5ab-2.4.6-ETCH-SDK92-i386.deb   jive5ab-2.4.6-LENNY-SDK92-i386.deb

2.4.5

This is a minor update, allowing jive5ab to be used with JIVE's version of SSErase - SSErase.py. SSErase.py depends on the Python "argparse" module (https://pypi.python.org/pypi/argparse (so it's available from pypi)).

SDK8.1 version: jive5ab-2.4.5-ETCH-SDK81-i386.deb

SDK8.2 version: jive5ab-2.4.5-ETCH-SDK8-i386.deb

SDK9.2 version: jive5ab-2.4.5-ETCH-SDK9-i386.deb   jive5ab-2.4.5-LENNY-SDK9-i386.deb

2.4.4

This is an IMPORTANT update, solely for Mark5B/B+ users, fixing a potential issue for geodetic/astrometric VLBI.

This fix deletes an implicit "synchronize clock to 1PPS" in the "record=on" and "in2net=on" commands. This implicit synchronization could cause clock jumps on the order of a few clock cycles (samples) between recorded scans, especially on systems where 1PPS and clock are not exactly synchronized.

The impact so far seems to have been minimal. This code has been in there for years and no (immediately obvious) problems have been observed in the analysis of astronomical (e-VLBI) or geodetic scans, but it would not hurt to be alert.

Other fixes in this update:

• FPDP2 mode is now always asserted on Mark5B+ at start-up, even if the hardware was programmed to use FPDP1
• at start-up the internal clock generator frequency is programmed identical to the clock frequency (see "clock_set="), with a maximum of 32 MHz
• proper detection of 2048 Mbps recording in "scan_check?" and "data_check?". Note: 2048 Mbps recordings have never been in danger, it was just the correct detection failing
• recording will not start if
  • there is no 1PPS signal selected [e.g. after a cold boot; between warm boots/restarts of the software the setting is remembered]
  • the selected 1PPS signal is not synchronized [synchronization is retained across warm boots and/or restarts of the software]
  • the DOT clock is not set up or running [the DOT clock is NOT retained across restarts of the software or any form of rebooting, necessitating "dot_set=" after a (re)start of the software]

SDK8.2 version: jive5ab-2.4.4-ETCH-SDK8-i386.deb

SDK9.2 version: jive5ab-2.4.4-ETCH-SDK9-i386.deb

2.4.3

This is a minor update to the 2.4.2 release only to actually fix the annoying and confusing "MARK5B FRAMENUMBER OUT OF RANGE" warning messages displayed during "scan_check?" or "data_check?".

SDK8.2 version: jive5ab-2.4.3-ETCH-SDK8-i386.deb

SDK9.2 version: jive5ab-2.4.3-ETCH-SDK9-i386.deb

2.4.2

UPDATE: 30 Oct 2013 User(s) reported that the advertised fix of "MARK5B FRAMENUMBER OUT OF RANGE" warning was, in fact, not fixed; the annoying warning message remains. The problem is being diagnosed and an update will be posted as soon as it is fixed.

This is a significant update to the 2.4.1 release. Annoying and confusing warning messages during "scan_check?" or "data_check?" have been removed.
2048Mbps Mark5B format is now properly diagnosed by "scan_check?" and "data_check?". Recording was always ok.
On Mark5B/DIM "clock_set" and DOT now work as intended; full support for DOT1PPS not running at 1PPS.

SDK8.2 version: jive5ab-2.4.2-ETCH-SDK8-i386.deb

SDK9.2 version: jive5ab-2.4.2-ETCH-SDK9-i386.deb

2.4.1

SDK8.2 version: jive5ab-2.4.1-ETCH-SDK8-i386.deb

SDK9.2 version: jive5ab-2.4.1-ETCH-SDK9-i386.deb

2.4.0

SDK8.2 version: jive5ab-2.4.0-ETCH-SDK8-i386.deb

SDK9.2 version: jive5ab-2.4.0-ETCH-SDK9-i386.deb

2.0.4

SDK8.2 version: jive5a-2.0.4-SDK8-i386.deb

SDK9.2 version: jive5a-2.0.4-SDK9-i386.deb

2.0.2

2.0

2.0-beta

1.2.3

1.2.0