A copy of libf2c.a (g77_5_21_libf2c.a) has been added to the lib directory and it should be possible to link in the new externals.
FMPUT. errors detected by FMVERI - entry will not be added to catalogueThe error is in FMUPTM, which is in packlib, and also affects the modules fatmen, fatnew and fatsrv.
This error can be made non-fatal by using
SET/VERIFY OFFor with the callable interface using
CALL FMSETV(0,IRC)
This bug is corrected since 17/18-May-96. kernlib, packlib and all modules affected were updated.
/bin/ld: Unsatisfied symbols: fpclassifyf (code) *** Error code 1This is not a bug in the libraries itsself. Hbook now has routines to check the validity of floating point numbers. The routines are provided by default only on PA1.1 systems ( all but hp710, hp720 ?), on other this must be done explicitely. The cernlib command does this; you may also add /lib/pa1.1/libm.a explicitly.
As has been discussed in a number of articles in the CERN computer newsletter over the last two years, we have switched to new code management and installation tools for the 96A release of the CERN Program Library.
This means that the CERNLIB source files are NO LONGER AVAILABLE in PATCHY or CMZ format (except in the OLD area.)
If you habitually use these files, e.g. to pick up "sequences", we would recommend that you move to using "#include files", as processed by the C pre-processor. Examples of how to do this will appear shortly on the CERNLIB web pages, accessible here
To ease the migration process, we will place copies of those "car" files that are needed in the /cern/pro/src/car area, containing only the sequence definitions. If you need one or more of these files, please send a mail to heplib.support@cern.ch stating which files you require.
Although it was not stated explicitly in CNL 207 in 1992, genlib in fact disappeared then and the CERNLIB command linked to mathlib instead. In addition, a link was made pointing to mathlib on all systems with the risk that it might sometimes be forgotten. In the belief that most users are using the cernlib command or explicitly referencing mathlib, we have decided to discontinue making this link. The use of the CERNLIB command is strongly recommended as it also provides other benefits, the most important being to platform specific system libraries (eg. X11) which are required by some Cernlib products. At CERN, in addition, the hepf77 command provides a portable interface to the native Fortran compilers, so we recommend for most applications, to link with-
hepf77 your.f `cernlib packlib mathlib kernlib`We are sorry for the inconvenience this is causing some users, but expect that removing the confusion about the existence and location of genlib will provide lasting benefits.
By mistake all routines of kernnum were compiled without optimisation on all systems. As an effect it was reported that some programs now take twice the time to complete. This is corrected on all systems (except SunOs) in kernlib/packlib since 12-June-96.
This bug is caused by a EPIO utility routine called ctoibm. This is not used by EPIO itsself. The bug is a compiler problem. The affected routine was recompiled by a "better" compiler.
This is fixed in packlib on SGI since 12-June-96.
This is a new routine for mathlib in 96a, announced in CNL 221.
This is fixed since 14-June-96 for all Unix, and since 21-June for VMS (both Alpha and VAX version).
This problem has now been fixed in the PRO areas at CERN. See hot news for more details.
Warning concerning the BUGREPORT command ======================================== Due to a bug in KUIP, the new command BUGREPORT doesn't work correctly in the PRO version of PAW. This problem will be fixed in the NEW area. Alternatively, continue to submit bug reports as previously.
Recommended solution is to use prescription described in CNL217 for transfer of RZ and other files.
Use the version of ZSERV from the NEW area. On CERNVM, add 'SETENV CERN_LEVEL NEW' to get the new version automatically. The problems were due to incompatible sequences between CSPACK and HBOOK.
Access the X11 libraries via "CERNLIB (X11"
A fix to these problems has now beem applied in the PRO area
A fix has now been applied to the PRO area on VXCERN (and the master source repository on asisftp.cern.ch)
Can be suppressed by compiling /NOWARN.
----------------------------------------------------------- * Give a short description of the bug * With several ntuples, when looping over ntuples with n/plot, there will be error messages ***** ERROR in HGNT : Bank does not exist : ID= 25 ***** ERROR in PPROJN : Error reading Ntuple : ID= 25 And the resulting histogram looks wrong. This worked correctly with the previous version of PAW.
Workaround: specify a valid pathname
*** For IBM/VM users ONLY *** In the new CERNLIB release (94B), all the versions of HIGZ (GRAFGKS, GRAFX11, GRAFGDDM) contain C code. In the previous release only GRAFX11 had C code. So, a CALL INITC() at the beginning of the main program is now mandatory for all the ^^^^^^^^^^^^ programs using GRAFLIB (HPLOT) or HIGZ (GRAFGKS, GRAFX11, GRAFGDDM) on IBM/VM. This call (a dummy C routine) forces the loading of the C library. This routine is in KERNLIB.