IOSTAT returns 37 on a write

rdtintel · ‎08-06-2012

Hi,

I am working with some Intel Fortan code that contains a line like:

write (UNIT=foo, REC=bar, IOSTAT=status) buff(1 : nbytes)

Very infrequently the status returned is 37. There does not appear to be anything strange

with the values of foo, bar, buff, or nbytes. However, "buff" is in a common block in shared

memory. This is a large system with interacting processes, something strange could be going on.

According to "Intel Fortran Building Applications", this is

Inconsistent record length

FOR$IOS_INCRECLEN. An attempt was made to open a direct acces file without

specifying a record length.

My question, is: What is the meaning of IOSTAT=37 for a write? Is it a normal return value from

write? When this happens, the file has usually been opened for quite a while.

If this is not a normal return value, then I have to consider the possibility that

the memory associated with file IO is being overwritten or something equally

difficult to debug. However, before going downthat road, i'd like to verify that

this IOSTAT value does not simply indicate a problem in the environment.

A pointer to the list of possible IOSTAT values for write would be great. I know where the list

of all possible IOSTAT return values is, but I don't know which ones are normally possible for write.

Thanks,

Ron

Steven_L_Intel1 · ‎08-06-2012

There is no list of possible error codes by statement. I could imagine this happening on a WRITE if unit FOO was not currently open. Either variable FOO changed its value or the unit got closed earlier in the program.