Coding to detect end of file or read error just won't work, and I can't see a
work around. Coding to differentiate between end of file and error condition
doesn't work, it all acts like an error.
I have a test code, included below, which creates a simple unformatted file
with two records. The first record contains a single integer, and the second
record contains three integers.
Once this test file is created, it is re-opened and read, attempting a read of
three integers, rather than 1 integer, on the first record. This should create
a error condition, and it should write out the stop error message stating so.
Instead, it writes out the message for the END=100 label.
Using IOSTAT doesn't differentiate between end= and err= conditions either, it
seems to return 0 for no problems, and -1 for anything else...
Code has been run on opteron, centrino, and itanium systems, all running linux
(fedora core 8 on the first two, last Suse)
!---Relevant Code follows----
integer ni, nj, nk
ni = 1
nj = 1
nk = 1
open(1,file='test.data',form='unformatted')
write(1) ni
write(1) ni,nj,nk
close(1)
ni = 0
nj = 0
nk = 0
open(1,file='test.data',form='unformatted')
read(1,err=10,end=100)ni,nj,nk
write(*,*)ni, nj, nk
stop 'No End, No ERror'
10 continue
write(*,*)ni, nj, nk
stop 'ERror'
100 continue
write(*,*)ni, nj, nk
stop 'End'
end
!-------------------End of Code ----------------
--
Summary: iostat and/or end= and Err= don't work in read
statements, unformatted reads
Product: gcc
Version: 4.1.2
Status: UNCONFIRMED
Severity: major
Priority: P3
Component: fortran
AssignedTo: unassigned at gcc dot gnu dot org
ReportedBy: michael dot e dot olsen at nasa dot gov
http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36942
