Smart to put "Flush(myfile)" before "Close(myfile") ?

Smart to put "Flush(myfile)" before "Close(myfile") ?

[reinert]

Can an exception (runtime error) in principle, after closing a file, interrupt the process of writing the actual file to disc?

I.e. is it smart to include a "Flush(myfile)" before "Close(myfile)" like this:

Create  (myfile, Out_File, "myfilename");
... Put .. Put
Flush(myfile);
Close(myfile);

reinert

Re: Smart to put "Flush(myfile)" before "Close(myfile") ?

[Jeffrey R. Carter]

On 1/25/21 6:42 AM, reinert wrote:
> Can an exception (runtime error) in principle, after closing a file, interrupt the process of writing the actual file to disc?
> 
> I.e. is it smart to include a "Flush(myfile)" before "Close(myfile)" like this:

In over 35 years of using Ada, I have never done this, and have never had a 
problem. YMMV

-- 
Jeff Carter
"You cheesy lot of second-hand electric donkey-bottom biters."
Monty Python & the Holy Grail
14

Re: Smart to put "Flush(myfile)" before "Close(myfile") ?

[Dmitry A. Kazakov]

On 2021-01-25 06:42, reinert wrote:
> Can an exception (runtime error) in principle, after closing a file, interrupt the process of writing the actual file to disc?

Depends on the error. If you crash the OS some cached stuff might get 
lost. That depends on the filesystem. If you managed to wipe out the 
RAID controller's cache as well as its persistent memory that would 
destroy anything there. Let your fantasy fly...

> I.e. is it smart to include a "Flush(myfile)" before "Close(myfile)" like this:
> 
> Create  (myfile, Out_File, "myfilename");
> ... Put .. Put
> Flush(myfile);
> Close(myfile);

AFAIK closing file flushes all file buffers and writes the file size. 
Thus calling Flush should change nothing.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de

Re: Smart to put "Flush(myfile)" before "Close(myfile") ?

[G.B.]

On 25.01.21 13:05, Dmitry A. Kazakov wrote:
> On 2021-01-25 06:42, reinert wrote:
>> Can an exception (runtime error) in principle, after closing a file, interrupt the process of writing the actual file to disc?
> 
> AFAIK closing file flushes all file buffers and writes the file size. Thus calling Flush should change nothing.

If the file is like standard output, say, then writing to its
physical target---there could be redirections etc.--might need
consideration when timing matters.
Also, on Windows™ I'd check the effects that the OS's buffering scheme
is having on synchronization of file contents and on timing.

I'm not sure whether the LRM reaches far enough for your case when it
mentions "internal buffers" and "external file". The latter, in
particular, seems out of reach when the file is read through means
other than Ada's.

Some older note-to-self I found states:
    -- When an I/O exception occurs, and a read-write loop would be
    -- left, it has helped (somewhat) to Flush the output stream
    -- (and possibly close the file).

Re: Smart to put "Flush(myfile)" before "Close(myfile") ?

[reinert]

OBS, me in a state of confusion: I made the original question since the end part of a file were lost (Flush helped to avoid).
However, I now discovered that the actual file sometimes were not closed :-) Anyway, you triggered me to find the bug :-)

reinert

Re: Smart to put "Flush(myfile)" before "Close(myfile") ?

[Dmitry A. Kazakov]

On 2021-01-26 06:30, reinert wrote:
> OBS, me in a state of confusion: I made the original question since the end part of a file were lost (Flush helped to avoid).
> However, I now discovered that the actual file sometimes were not closed :-)

This is what Flush is for. E.g. if you have some trace or log file which 
never gets officially closed. Flush sets the file end.

  Anyway, you triggered me to find the bug :-)

You can wrap File_Type in a controlled type to ensure Close called from 
the Finalize.

-- 
Regards,
Dmitry A. Kazakov
http://www.dmitry-kazakov.de