[Info-vax] Notable VMS "feature"

[Johnny Billquist]

On 2016-01-28 16:02, VAXman- at SendSpamHere.ORG wrote:
> In article <n8d3v9$8l3$1 at Iltempo.Update.UU.SE>, Johnny Billquist <bqt at softjar.se> writes:
>> On 2016-01-28 03:18, VAXman- at SendSpamHere.ORG wrote:
>>> In article <n8bj49$u3u$1 at Iltempo.Update.UU.SE>, Johnny Billquist <bqt at softjar.se> writes:
>>>> On 2016-01-27 12:54, VAXman- at SendSpamHere.ORG wrote:
>>>>> In article <n89ec2$1bpl$1 at gioia.aioe.org>, George Cornelius <cornelius at charter.NOWHERE.net> writes:
>>>>>> On 01/26/2016 06:31 AM, Jess Goodman wrote:
>>>>>>> On Sunday, January 24, 2016 at 3:40:18 PM UTC-5, gcorn... at charter.net wrote:
>>>>>>>> gcornelius at charter.net wrote:
>>>>>>>>> gcornelius at charter.net wrote:
>>>>>>>>>> The rest is of course stolen from you.
>>>>>>>>
>>>>>>>> Except the part where I forgot the parentheses
>>>>>>>> about istat.and.1 in the if statement.
>>>>>>>>
>>>>>>>> I believe the simplified form
>>>>>>>>
>>>>>>>> 	if (istat.and.1)  ! no .eq.1
>>>>>>>>
>>>>>>>> would work, though, so no need to have compared
>>>>>>>> the results of the .and. to 1.
>>>>>>>>
>>>>>>>> George
>>>>>>
>>>>>>> Actually in VMS Fortran (and in DCL) only the low-order bit of a
>>>>>>> logical value determines if it is TRUE or FALSE.  I remember this
>>>>>>> as "truth is odd".  And since for VMS status values "success is
>>>>>>> odd" you can code this way:
>>>>>>
>>>>>> And I code that way in DCL. In Fortran, which is
>>>>>> a standards-based language, I have begun to think
>>>>>> that even if it is low bit only it is best to make
>>>>>> it explicit.
>>>>>
>>>>> Then wouldn't be be more "explicit) with...
>>>>>
>>>>>            INCLUDE '($STSDEF)'
>>>>>
>>>>>            if (istat.and.STS$M_SUCCESS)
>>>>
>>>> Isn't there a mask that contains all the bits that should be checked for
>>>> status? And then you should really mask istat with that, and check if
>>>> the masked value equals success?
>>>
>>> "M" ;)
>>
>> I was wondering about that, but is STS$M_SUCCESS is a mask, your code
>> don't make sense, as errors would also result in a non-zero value. Or is
>> STS$M_SUCCESS a mask that only includes the bits that indicates
>> success/no success?
>>
>> Hmm, I bet it is. So I would guess that it's a mask with just one bit?
>>
>> Oh well. In which case I agree with other comments still, it's better to
>> still check that the result value actually equals success, than just
>> assuming that the success value is actually when it is non-zero, and
>> that any non-zero value from an expression is interpreted as .TRUE.
>>
>>
>> Making up names, I would prefer something like:
>>
>> IF ((istat .AND. STS$M_STATUS) .EQ. STS$C_SUCCESS)
>
>
> IF ((istat .AND. STS$M_SEVERITY) .EQ. STS$K_SUCCESS) ...
>
> Howver, in that case, you'd be assuming that in information code was an
> error unless you check for that too.
>
> IF (((istat .AND. STS$M_SEVERITY) .EQ. STS$K_SUCCESS) .OR.
>      ((istat .AND. STS$M_SEVERITY) .EQ. STS$K_INFO) ...

Thanks, and yes, indeed. Good point. You want to check for more than 
just success.

Or, I guess, you could do:

IF ((istat .AND. STS$M_SUCCESS) .EQ. STS$K_SUCCESS) ...

which should cover it all.

	Johnny

-- 
Johnny Billquist                  || "I'm on a bus
                                   ||  on a psychedelic trip
email: bqt at softjar.se             ||  Reading murder books
pdp is alive!                     ||  tryin' to stay hip" - B. Idol