Correction Author: David A. Pearlman
Correction Date: 10/13/91

Programs: sander

Severity: Moderate

Problem:  If time-averaged restraints are requested in SANDER using 
     a *namelist* type=DISAVE/ANGAVE/TORAVE weight change instruction, 
     and, additionally, the "portable" namelist code is being used, 
     the time-averaged restraint card will be ignored with the message:

           "Warning: Averaging request ignored".

     This message will appear even when the time-averaged request 
     instruction is valid. (N.B. Such requests are correctly ignored when 
     minimization is being performed). 

     This problem would NOT arise if you either specified the "weight
     change" instructions in formatted form (&formwt), or if you
     compiled sander with the intrinsic namelist for your host machine.
     (See the file amber4/src/MACHINE; If -DREGNML appears appears after 
     the phrase "setenv MACHINEFLAGS", you are using the machine
     intrinsic namelist. If -DREGNML does not appear, you can add it,
     and use the intrinsic namelist if it exists [most, though not
     all, machines now support an intrinsic namelist].

Affects: Time-averaged restraints requests will be ignored under the
     conditions described above.

Cause: A single integer variable related to the portable namelist read
     was not being correctly initialized in routine RESTAL. This routine
     does a cursory read of the restraints file to determine how much
     memory needs to be allocated for restraints. The mis-initialization
     resulted in only the first four characters of the string provided 
     for the TYPE= field being stored. Consequently, no match for the 
     >4 character time-averaging request strings would be found in 
     RESTAL, no memory would be allocated for time-averaged restraints, 
     and subsequently, any time-averaged restraint requests would be ignored.

Fix: In routine RESTAL (file amber4/src/sander/restal.f), make the
     following changes:

        old lines:    data attnm2/5,1,1,3,3,1,1/
                      data attnm2/5,1,1,2,2,1,1/

        new lines:    data attnm2/3,1,1,3,3,1,1/
                      data attnm2/3,1,1,2,2,1,1/

     (only change in each data statement is the initial 5 should be a 3).

Temporary workarounds: The fix is simple. But if recompiling is 
     inconvenient, and the program was compiled with the "portable"
     namelist, simply specify the weight change cards in formatted form,
     (&formwt) as described in the manual.


Routines affected:

     SANDER:  Routine RESTAL in file .../amber4/src/sander/restal.f