% README File FOR AIPS Release 15APR91 % This File describes the patches to known bugs in 15APR91 Tasks for which patches have been made: Task: VTESS Bug: Did not make 4K by 4K Images, because a buffer was too small. File(s) 15APR91/QY/PGM/NOTST/VTESS.FOR Fix: Compile and Link new Version of VTESS.FOR COMLNK $QYPGNOT/VTESS.FOR Documentor: Glen Langston, NRAO C'ville (804) 296-0328. Date: 1991 July 26 Task: INSTEP4 Bug: Does not rebuild all the AIPNOT tasks on Unix install. This is not serious enough to break the installation, as INSTEP3 will explicitly compile/link the vital tasks from this area. However, there are some useful utility programs to be found there. File(s): 15APR91/SYSTEM/UNIX/INSTALL/INSTEP4 Fix: Add AIPNOT to the list of areas processed by INSTEP4 - or get the replacement INSTEP4 from SYSTEM/UNIX/INSTALL here - or just do a "COMLNK $AIPNOT/*.FOR". Documentor: Pat Murphy, NRAO/CV. Date: 1991 Aug 19 Task: All (SunOS with Sun Fortran 1.3 through 1.4) Bug: Interaction of TMPFS kernel option with compilation. There is a bug in certain versions of Sun Fortran that causes some data statements to be initialized incorrectly. This bug is only active when the SunOS TMPFS kernel option is enabled and the "/tmp" area is mounted on the swap partition. It also only seems to affect data statements where the first item is to be a zero. The usual symptom is a series of messages complaining about "LUN 12 (or 6) ALREADY CLOSED (or open) IN FTAB". File(s): All Fortran modules with DATA statements Fix: The following fixes may work with varying degrees of effectiveness; only the first two are guaranteed to work. 1) Apply the Sun patches 100174-01 (and 100175?) or 2) remake your kernel with the TMPFS line commented out from the config file, and mount /tmp on its own partition. Then rebuild all of AIPS. or 3) Change the COMP= line in $SYSLOCAL/FCOPTS.SH and CCOPTS.SH to include "-temp=$HOME/tmp" or something like this, i.e. force fortran to use a different area for temporary file placement. or 4) don't mount /tmp on the swap partition. This approach has NOT been tested. Let us know if this works! Documentor: Pat Murphy, NRAO/CV. Date: 1991 Aug 19; updated 1991 Dec 03. Tasks: MX and HORUS Bug: When making 1024x1024 (or larger) images of spectral line data using UNIFORM weighting. All channels after the first were incorrectly weighted, creating something which was similar to, but slightly worse than, NATURAL Weighting. (Note: there were NO problems with Natural Weighting or Images smaller than 1024x1024.) Bug was found by Ralph Gaume of the Naval Research Lab. File: 15APR91/Q/SUB/NOTST/UVTBUN.FOR Fix: Grid used for UNIFORM Weighting was not initialized each execution, and for large images, sections already contained values which corrupted the weighting. Fix: Get new copy of UVTBUN.FOR, compile UVTBUN and relink tasks MX and HORUS. COMRPL $QNOT/UVTBUN.FOR COMLNK $QYPGNOT/MX.FOR COMLNK $QPGNOT/HORUS.FOR Documentor: Glen Langston, NRAO C'ville (804) 296-0328 Date 1991 October 15 Task: CALIB Bug: Was not using more than 1000 clean components. If more than 1000 were supplied, they were not used. Bug found by Frazer Owen at the VLA. File: 15APR91/Q/PGM/NOTST/CALIB.FOR Fix: Array size was increased. Get new version of CALIB and compile and link it. COMLNK $QPGNOT/CALIB.FOR Documentor: Glen Langston, NRAO C'ville (804) 296-0328 Date: 1991 October 15. Runfile: MAPIT.001 Bug(s); Overwrote ADVERB FLUX, did not allow user to look at intermediate images, no interactive mode. Major improvements in latest version, including ability to restart self-calibration process. Files: 15APR91/RUN/MAPIT.001 15APR91/HELP/MAPIT.HLP 15APR91/HELP/MAPIT_NA.HLP 15APR91/HELP/MAPIT_UV.HLP 15APR91/HELP/MAPIT_MX.HLP Fix: Get new RUNFIL containing procedures, place it in the AIPS $RUNSYS area. Place help files in the $HLPFIL directory. Documentor: Glen Langston, NRAO C'ville (804) 296-0328 Date: 1991 November 25. Task: CLIP Bug: Failed clipping 256 channel Spectral Line data because a buffer was too small. Program returned an error code and dies when encountering a file requiring too large a buffer. File: 15APR91/APL/PGM/CLIP.FOR Fix: Array size was increased. Get new version of CLIP and compile and link it. COMLNK $APLPGM/CLIP.FOR Documentor: Glen Langston, NRAO C'ville (804) 296-0328 Date: 1991 December 10. Task: FILLM Bugs: SUBTLE PROBLEM #1. The reference frequency was changed with each new scan. Thus the (u,v,w) were not, in general, referred to the reference frequency in the catalog header! Frazer Owen is credited with finding this effect, which is now fixed. SUBTLE PROBLEM #2. When changing UV files, the CL table entry at the beginning of the next scan was not written! This can lead to some significant problems with applying calibration. Both of these problems are, of course, correctable using UVFIX and INDXR, respectively; but how many people did that? 3. The total bandwidth were incorrectly determined for the narrowest back-end filters/recirculators. 4. During July and August 1990, the VLA IAT clock was misbehaving. Unfortunately, this was the first test of the on-line software to handle clock errors; the software also failed. This led to total delays of the order of 1.0D50 nanoseconds being recorded on the VLA archive tape. Processing of these large numbers caused FILLM to die on Convexes and to produce nonsense results on Suns. Out-of-range delays now cause all data to be flagged bad. File: 15APR91/APL/PGM/NOTST/FILLM.FOR Fix: Get new version of FILLM.FOR COMLNK $APGNOT/FILLM.FOR Documentor: Gareth Hunt, NRAO, Charlottesville (804)-296-0236 Date: 1992 February 21 Task: None -- Dynamic Libraries for AIPS on SUN4, SunOS 4.1.1 Bug: None. Files: 15APR91/SYSTEM/UNIX/SUN/FCOPTS.SH 15APR91/SYSTEM/UNIX/SUN/CCOPTS.SH 15APR91/SYSTEM/UNIX/SUN/LIBR 15APR91/SYSTEM/UNIX/SUN/LINK Fix: To enable the use of shared libraries, copy these files to your $SYSLOCAL area on your SUN installation, possibly editing the FCOPTS and CCOPTS files for optimization or debug settings (but LEAVE the -PIC and -pic qualifiers as they are!!). Then cd to $INSUNIX, remove any *.LOG and *.LIS files, and rerun INSTEP2 and INSTEP4. This reduces the binaries (contents of the $LOAD area) from over 200 megabytes to around 60 megabytes. Documentor: Pat Murphy, NRAO-Charlottesville, (804) 296-0372, Date: 1992 March 13 (Friday!) Task: SPLIT Bug: Cannot SPLIT data calibrated in versions of AIPS 15OCT92 and NEWER. CL table format in new versions not recognized in 15APR91 AIPS. File: 15APR91/APL/SUB/NOTST/CLREFM.OR Fix: CL table reformating subroutined fixed. Get new subroutine CLREFM.FOR Compile subroutine: COMRPL $APLNOT/CLREFM.FOR Relink task: COMLNK $APGNOT/SPLIT.FOR Documentor: Glen Langston, NRAO C'ville (804) 296-0328 Date: 1992 October 5 ----------------------------------------------------------------------- Task: FITTP, VLBIN, TVMON Summary: Antenna tables written by FITTP are bad on DecSystems. Bugs: For DECSTATIONS and DECSYSTEMS running ULTRIX only. (actually for any computer using IEEE floating point format that has little endian byte order). There is a bug in the ZRLR64 subroutine that can affect antenna tables and any other double precision output on FITS tape or disk file. Bytes were not getting swapped. The modified ZRLR64.FOR routine here fixes the problem. Files: /15APR91/APL/DEV/ZRLR64.FOR Fix: (1) Get ZRLR64.FOR and place in $APLGEN (2) Recompile: COMRPL $APLGEN/ZRLR64.FOR (3) Rebuild tasks: COMLNK $APGNOT/FITTP $APGNOT/VLBIN $YPGVDEV/TVMON Documentor: Pat Murphy (Bug found by JingPing Ge at Brandeis). Date: 1992.11.06 -----------------------------------------------------------------------