What is severe error 38?

[eaengqu]

We have your Intel 10 compilers here at Sandia. One of our users is getting this message:

severe (38). Error during write FOR$IOS_ERRDURWRI. The Compaq Fortran RTL I/O system detected an error condition during execution of a WRITE statement.

[Steven_L_Intel1]

I'm a little confused here - if you are using the Intel compiler, why are you referencing Compaq Fortran? Also, I would not expect to see a message exactly like that displayed; the "Error during write" is fine, but the rest looks like it was text copied from a manual.

In any case, the answer is the same. When the Fortran run-time library called the operating system to do a write operation, the OS returned an unexpected error. This is usually due to a problem in the environment (network share disconnected, disk full, etc.)

Was there no additional information displayed? Usually there is, unless the program captured the error and displayed the message itself.

[eaengqu]

Yeah the message with Compaq in it made me wonder. I was wondering if we had a bad rtl, or whatever. The user is our case is running VASP on a parallel Linux cluster with Myrinet and Openmpi. We are just looking into the problem now . I
checked the directory for the I18N messages (which is
/projects/global/x86_64/compilers/intel/intel-10.0-fce-025/lib/locale/en_US and still see error messages with Compaq references:

P.S. If you work with Ron Green in Albuquerque/Rio Rancho, he knows us and or machine for when he was a Compaq employee.

$ foreach i (`ls`)
foreach? echo $i
foreach? strings $i | grep ompaq
foreach? end
diagsf.cat
Error occurred during /ANALYSIS_DATA processing. Please report the problem Compaq.
This beta test software has expired. Please contact Compaq to purchase a released version.
diagsf.msg
diag_F90_FFE_SEMANTIC_1754 Error occurred during /ANALYSIS_DATA processing. Please report the problem Compaq.
diag_F90_ME_TIMEBOMB This beta test software has expired. Please contact Compaq to purchase a released version.
ifcore_msg.cat
ifcore_msg.msg
irc_msg.cat
irc_msg.msg
libm.cat
libm.msg

[Steven_L_Intel1]

Fascinating, though I don't see the text for error 38 in there. The messages you cite would not ever be emitted by the Intel compiler. (We still share sources with the Compaq/HP OpenVMS Itanium
compiler in some ways.)

In any case, there is some OS error that occurred that you'll have to identify. If you know which WRITE statement had the problem, you can call ERRSNS after it and see if the system error code is provided.