schell@iris.ucdavis.edu (Stephan Schell) (06/14/89)
I have a problem using namelists on Hewlett-Packard's 9000/825 under HP-UX
3.0 that I have not encountered on other machines (e.g., VAXen).
HP-UX 3.0 (9000/825 machine) version of f77 does not seem to accept namelist
input like (I think) it should. I have a real array in the namelist. If I
do not specify each and every element's value in the namelist input file,
then I get a nasty error message. If I do specify every element's value,
then all is well.
The HP-UX f77 manual seems to indicate that what I'm doing should be OK.
I don't think the user of the program should be responsible for filling
up array elements that he does not want to use.
OS release and version info: HP-UX A.B3.10 B
Here is file "test.f". Typing "f77 -a -K -o test test.f" produces
no warnings. (check for anything that is non-ANSI according to HP).
PROGRAM TEST
CHARACTER*16 A(3)
REAL B(3), C(3), D(3)
NAMELIST /FOO/ A, B, C, D
OPEN(UNIT=10,FILE='TEST.INP',ACCESS='SEQUENTIAL',STATUS='OLD',
1 FORM='FORMATTED')
READ(UNIT=10,NML=FOO)
CLOSE(UNIT=10)
STOP
END
Here is file "TEST.INP". Notice that only the first element of each array
is given a value.
$FOO
A = 'AZ',
B = 0.0,
C = 360.0,
D = 0.1
$END
Here is the result of typing "test".
*** FORTRAN I/O ERROR 907: ERROR IN LIST-DIRECTED I/O READ OF CHARACTER DATA
( 0) 0x00007128 IO_DED + 0x250
( 1) 0x00004d14 FTN_F_RSNE + 0x154
( 2) 0x00002ed0 test + 0x128
So, what did I do wrong, or have I found a bug, or are the namelist standards
too vague to make a determination here?
Thanks for anything you have to say.
Stephan V. Schell
Dept. of Electrical Engineering
University of California, Davis
schell@iris.ucdavis.edu
...ucbvax!ucdavis!iris!schellkhb@chiba.Sun.COM (Keith Bierman - SPD Languages Marketing -- MTS) (06/14/89)
In article <4654@ucdavis.ucdavis.edu> schell@iris.ucdavis.edu (Stephan Schell) writes: >I have a problem using namelists on Hewlett-Packard's 9000/825 under HP-UX >3.0 that I have not encountered on other machines (e.g., VAXen). ... > >OS release and version info: HP-UX A.B3.10 B >Here is file "test.f". Typing "f77 -a -K -o test test.f" produces >no warnings. (check for anything that is non-ANSI according to HP). Well, NAMELIST is not in x3.9-1978 (aka fortran 77) so file a bug against the non-ansi detection :> > >Here is the result of typing "test". >*** FORTRAN I/O ERROR 907: ERROR IN LIST-DIRECTED I/O READ OF CHARACTER DATA >( 0) 0x00007128 IO_DED + 0x250 >( 1) 0x00004d14 FTN_F_RSNE + 0x154 >( 2) 0x00002ed0 test + 0x128 > >So, what did I do wrong, or have I found a bug, or are the namelist standards >too vague to make a determination here? There is no NAMELIST in the "old" standard. so anything a vendor does is "ok". Your code happens to run on any sensible namelist implementation (vax, sun, lahey, univac for starters). So I would file a bug report. Keith H. Bierman |*My thoughts are my own. Only my work belongs to Sun* It's Not My Fault | Marketing Technical Specialist ! kbierman@sun.com I Voted for Bill & | Languages and Performance Tools. Opus (* strange as it may seem, I do more engineering now *)
milburn@me10.lbl.gov (John Milburn) (06/14/89)
In article <4654@ucdavis.ucdavis.edu> schell@iris.ucdavis.edu (Stephan Schell) writes: >I have a problem using namelists on Hewlett-Packard's 9000/825 under HP-UX >3.0 that I have not encountered on other machines (e.g., VAXen). > >HP-UX 3.0 (9000/825 machine) version of f77 does not seem to accept namelist >input like (I think) it should. I have a real array in the namelist. If I [...] > $FOO > A = 'AZ', > B = 0.0, > C = 360.0, > D = 0.1 > $END > I had the same problem, until I realized that the hp-ux namelist, unlike VAX/VMS namelist, would like you to give it " &" as a record delimiter, rather than "$". Since then, have had no trouble. __________________________________________________________________________ | John Milburn - Advanced Light Source - Lawrence Berkeley Laboratory | | INTERNET: JEMilburn@lbl.gov BITNET: JEMilburn@LBL.bitnet | | DECNet: LBL::JEMilburn UUCP: {...}!ucbvax!lbl.gov!JEMilburn| | Ma Bell: (415) 486-6969 SnailMail: 1 Cyclotron Road 46-161 | | Facsimile: (415) 486-4873 Berkeley, Ca. 94720 | --------------------------------------------------------------------------
cdb@hpclcdb.HP.COM (Carl Burch) (06/25/89)
> Well, NAMELIST is not in x3.9-1978 (aka fortran 77) so file a bug > against the non-ansi detection :> I hate it when the competition's right. :-) Keith's correct, the non-ANSI detection was left off the NAMELIST implementation by accident. On the original run-time problem, it will go away if you remove the white space around the '=' (i.e., "C=360.0," instead of "C = 360.0,"). As often happens, the fix for another problem caused this one - we forgot to allow whitespace (tabs, spaces) after the variable name that terminates the proceding array list. I suspect the compiler team will correct both of these problems for the next release, now that they have been pointed out. By the way, both '$' and '&' delimiters for namelist data should be accepted (they are on this example). If you can find an example where one will work and not the other, please e-mail me. - Carl Burch