sscan Release Notes

Release 2-6

• The sscan record can now post current-data arrays during a scan. While ATIME >= 0.1, ALL arrays will be posted when a new data point has been acquired and ATIME seconds have elapsed since the last array posting. New sets of array PV's have been added for this purpose, since the old array PV's must contain the previous scan's data to avoid breaking data-storage clients. The new PV's are PnCA (positioners, e.g., P1CA), and DnnCA (detectors, e.g., D01CA). During a scan, arrays are posted with the attribute DBE_VALUE; at end of scan, they are posted with DBE_LOG as well.

  Unfortunately, posting current-data arrays during a scan results unavoidably in the posting of the previous-data arrays, PnRA and DnnDA. Clients that monitor these PV's can regain their old behavior by specifying the mask DBE_LOG in their ca_add_event() or ca_create_subscription() call.

• The MEDM display scanDetPlot.adl now uses the new current-data PV's to display data. (These PV's also get end-of-scan data.) The MEDM display scanDetPlotRT.adl has been renamed scanDetPlotFromScalars.adl.
• Previously, the sscan record repeated final data values out to the ends of arrays, when a scan was finished, to aid display clients that don't know how to plot only a PV-specified number of data points. Now this treatment can be done also during a scan, as controlled by the PV COPYTO.

Release 2-5-7

• Allow end user to specify the base name of data files written by saveData. Previously, the ioc prefix (modified to avoid characters illegal in file names on the supported operating systems) was used as the base file name. Now, if saveData.req contains the section [basename], and the PV named in that section exists, and the string value of that PV is not the empty string, then saveData will use the string value, instead of the ioc prefix, as the base file name (onto which the scan number and the string ".mda" will be appended).
• Previously, saveData's init file could not usefully specify PV names containing the characters '-', '[', ']', '<', '>', or ';', even though these are legal characters for a PV name.
• New documentation files: saveData.req and scanParmRecord.html
• Busy record now pays attention to it's UDF and alarm fields, executes its its DOL link only if that link is not CONSTANT.
• If recDynLink encounters a link structure that thinks it has an instance on queue, but the queue is in fact empty, then the link structure is corrected.
• saveData's stack size increased.to epicsThreadStackBig.

Release 2-5-6

• scan.db database separated into standardScans.db and saveData.db.
• Added standardScans_settings.req and sscanRecord_settings.req. This allows a script to more easily write a new auto_settings.req file, since the request file has the same name as the database it supports. Also, this makes it easy to load more than one copy of standardScans.db.
• Win32-specific .dbd file is no longer needed, since Mark Rivers added saveDataWin32.c, which contains stub functions for commands that could not be built for Win32.
• saveData now checks all data-file writes for errors, and retries until file is successfully written, or user-specified number of retries has been done. User also specifies the time between retries. The new PV's were added to scan_saveData.adl
• In sscan module 2.5.3, saveData was writing scan-dimensions to the wrong file offset, under certain circumstances. This is fixed.
• recDynLink now calls epicsAtExit, so it can avoid making CA calls after CA has been shut down.
• recDynLink handles null and empty PV names more gracefully.
• sscan record now has a CMND-field value for clearing positioner drive and readback PV's, and the default medm display file uses this value for it's "CLEAR" button.

Release 2-5-3

• Added sscanApp/op/python directory, with the following programs:

  addMDA.py

  Front end for adding MDA files, uses readMDA, opMDA, and writeMDA from mda.py

  mda.py

  Python API for MDA files. Supports reading, writing, and arithmetic operations for up to 4-dimensional MDA files

  mdaAsc.py

  Uses mda.py to render a 1-dimensional MDA file as ascii text.

  opMDA.py

  Front end for operating on MDA files, uses readMDA, opMDA, and writeMDA from mda.py

• Fixed problems in the communication between the sscan record and saveData that caused corrupted data files to be written:
  • The basic problem was that saveData was getting bufferred data arrays, but an unbuffered copy of the sscan record's CPT field. The sscan record now maintains the field BCPT (bufferred CPT) which is posted when data array buffers are switched.
  • A second problem was that saveData was not able to put AWAIT=1 quickly enough to stop a very fast scan in time to ensure integrity of the data file. saveData now writes '1' to the sscan record's AAWAIT field on init, and writes '0' if it ever exits (not a supported operation at this time). As a consequence, AAWAIT no longer occurs in the autosave-request file scan_settings.req.
  • A remaining problem, thus far seen only on cygwin, is that multidimensional scans can get saveData into trouble because CA monitors sometimes are received by saveData in a different order than they were posted by the sscan records. Currently, neither the sscan record nor saveData defend against this.
• Added Dohn Arms' 'mdautils' software in the sscanApp/src directory. This software can convert an MDA file to ascii, print info about an MDA file, and read an MDA file into C data structures.
• Fixed a race condition in the sscan record that was responsible for hanging scans at the last point (and maybe other things as well).
• the sscan record no longer renews PV links when a scan starts if the new scan follows the previous scan by less than sscanRecordLookupTime.
• If retrace or after-scan fails because recDynLinkPutCallback returns an error, skip the action rather than hang.
• If the sscan record attempts to connect to a PV while an earlier connection attempt is still in progress, it now waits and retries.
• recDynLinkQsize is now exported for use by the ioc shell.
• recDynLink used to crash if one of its callback functions received an event_handler_args structure with a status element whose value was not == ECA_NORMAL. Now it declines to process the event or to pass it on to the client.
• saveData used to check directory permissions by attempting to create a file whose name was illegal (contained ':') on some operating systems.
• rewrote sscanRecord.html

Release 2-5-2

• sscanRecord checks parameters more closely, allows before-scan and after-scan links to write to selected PV's of their own sscan record.
• New after-scan action: Move to center of mass of peak (this choice has problems with multiple positioners, since they won't, in general, have the same peak position).
• In previous versions, recDynLink would deadlock if asked to clear the link to a PV while an action for that PV was still on queue. This is fixed.
• saveData zeros unused points in its XDR buffer, because XDR doesn't manage this well.

Release 2-4

• The sscan record and saveData now take advantage of the sscan record's double buffered data arrays, and allow a scan to proceed while the previous scan's data is being written to disk. (AWAIT, AAWAIT fields)
• saveData now runs on Solaris and Linux ioc's.
• Array valued detectors are now supported in all scan modes. Arrays are read at the end of the scan. If processing is required to get array PV's ready, the sscan record can trigger that processing with the A1PV, A1CD fields (just like detector triggers, but executed after the scan is done, and just before array PV's are acquired).
• The new field DSTATE shows the state of the sscan record's data arrays. When DSTATE==POSTED, the sscan record can begin a new scan.
• Previously, the before-scan and after-scan links always waited for completion. Now the user decides, by setting BSWAIT and ASWAIT.
• If scan fails limit checks, the scan hangs until user aborts. Previously, the scan would appear to complete.
• new medm displays for scans are simpler, less cluttered, and have documentation callups.
• In saveData.req (the file in the ioc directory that tells saveData what scans to monitor, etc), the handshake PV is now ignored. saveData now uses scanX.AWAIT to handshake with scanX.
• saveData no longer allocates local storage for unused sscan-record detectors. Once a sscan-detector PV is specified, storage is allocated and never released.
• Modified the scanparm record to support multi-dimensional, multiple- positioner scans: Added two output links to the scanparmRecord: OLOAD, and OGO. If LOAD is nonzero, it's written to OLOAD; if GO is nonzero, it's written to OGO.
• Added scanParms2Pos.db, scanParms2Pos.adl -- scan parameters for a 1D scan with two positioners.
• The scanparm record now uses long int, rather than short int, for the number of data points (fields MP and NP).
• sscanRecord.dbd and scanparmRecord.dbd now include sscanMenu.dbd, to ensure that menus are consistent. Sometime in the past, the scanparm record wasn't updated when sscanRecord menu fields were modified. In particular, scanParm*.db and alignParms.db used to specify "Relative" positioning, while the sscan record accepted only "RELATIVE" or "ABSOLUTE".
• The sscan record now issues recDynLinkGetCallback() for each positioner and detector PV, and waits for the callback before using the PV value that was cached by recDynLink.
• Added recDynLinkGetCallback() to recDynLink library. Also fixed some bugs.
• cvs log

Release 2-3

This is the first release of the synApps sscan module. Version numbering for this module begins with 2.3 because this module was split from version 2.2 of the std module, and I wanted to retain the CVS histories of module contents.

• Converted to EPICS 3.14. Currently saveData runs on vxWorks only.

• Docs updated and moved to sscan/documentation

• saveData - added iocsh support; changed number of data points from short to long int, to support very large scans. The data file format is unchanged, however, because the number of points was already being written as a four-byte quantity.

• sscanRecord - Number of points in a scan is essentially limited only by available memory. save-restored value of NPTS is now checked against MPTS. Array mode (ACQT="1D ARRAY") was broken. (The change from ACQM="ARRAYS" to ACQT="1D ARRAY" wasn't done correctly.)

  Previously, the sscan record's response to an abort request (.EXSC=0) while no scan was in progress (.BUSY==0) was to return nonzero from special(), and EPICS tolerated this without comment. Now it signals an error to the client. But we don't (always?) want this action to be regarded as an error. For now, the scan database just declines to abort a sscan record that isn't busy, but clients writing directly to the sscan record directly can still get this error message.

• recDynLink - Fixed memory leak (epicsMutex created but not destroyed). Switched communication with link tasks from ring buffer to message queue. recDynOut was calling ca_pend_event, which used to flush the ca buffer, but evidently no longer does; replaced with ca_flush_io.

• saveData_settings.req - new file.

• scan_settings.req - added fields ACQT and ACQM.