Re: Send_File API versus Send API over a Socket

[Scott Klement <rpg400-l@xxxxxxxxxxxxxxxx>]

Shannon,

The fgets() API (i.e. _C_IFS_fgets for IFS files) is designed for 
reading TEXT files.  Hopefully you understand the difference between 
text data and binary data -- if not, please ask and I'll explain.

Since an image file is NOT text data, these APIs are not suitable for 
reading it.

Either use the fread() (err... _C_IFS_fread) API, to read the file (and 
if you do, make sure you use the 'b' flag on fopen() to tell the system 
to read in binary mode, as well).

Or change your code to call the system IFS APIs open() and read() to 
read the file (instead of the ILE C fopen() and fread() APIs).  If you 
do use this approach, please make sure that you do NOT specify 
O_TEXTDATA -- because, again -- image files are not text data.

My suspicion of your problem is that you're either mistranslating the 
data (i.e. translating it from ASCII to EBCDIC with fgets() and then 
manually translating it back) and therefore corrupting the data 
(translation from EBCDIC to ASCII and vice-versa makes no sense on 
binary data.)   OR your image file contains nulls (x'00') which would 
make it impossible to determine the length of the data when using 
fgets().... making it likely that you're simply discarding any bytes 
after any x'00' in each block.

send_file() is the equivalent of calling read() and send() in a loop. 
You still have to call the open(), socket(), and connect() APIs 
manually, and you still have to remember to exclude O_TEXTDATA from the 
 open() call.    send_data() can also close() the socket when it's 
done.... but you still have to call close() on the IFS file.

Frankly, I never use send_data() for the simple reason that the 
operating system takes control during the entire send of the file.  This 
deprives me of the ability to show a "progress meter" to the user that 
details how much of the file has been sent, and how much is remaining, 
and other good stuff like that.   Since it's not difficult to code a 
loop between read() and send(), I just go ahead and code the loop. 
Send_File() really doesn't save me much work.

I'm almost certain that your problem has nothing to do with the 
difference between send_file() and send() -- but rather that you're 
reading the IFS file as if it's a text file, when it's clearly not!



sodonnell@xxxxxxxxxxxxxxxxxxxxxxxxx wrote:
> Hello,
>
> I have an application that reads an image file off of the IFS (it's a TIFF file if that makes a difference...) and then I send it to a PC over a 
> socket. 
>
> I'm using the _C_IFS_OPEN and _C_IFS_GETS to read the file and I'm using the Unix-type API Send to send the file across the socket.
>
> The problem I'm having is that somewhere along the way, my image file is getting corrupted.  That is, the first few lines of the image data are fine, 
> and in fact, you can even make out a little bit of what the image is if you look at it in Windows. But somewhere between the last point where I have 
> good data and the end of the file (or possibly including the end of the file too), my image data gets altered.
>
> If I look at the file in Notepad or Wordpad or even in a DOS prompt, I can see that the data is MOSTLY correct. however, there are places where the 
> data is obviously different. And since it's in hex (or whatever...) it's impossible for me to look at it and say, "Oh, there's a line feed or carriage 
> return there where there shouldn't be."
>
> So, what I'm wondering is... If I used the Send_File API (instead of SEND and the _OPEN and _GETS APIs (as shown above)), do you think my chances of 
> avoiding trashed data would improve?   I guess I could just try it and see but I was hoping someone might have used Send_File before and had some 
> experience with it.
>
> There are 3 distinct points where the data could be getting hosed in this process:
>
> 1. When I read the data using the _C_IFS_GETS API.
> 2. When the data is sent across the socket from my RPGLE program.
> 3. When the data is received at the socket server.
>
> The last one does not seem as likely simply because that process has worked flawlessly for years.  On the other hand, I don't know if I've ever send 
> an image file like this before to it. This might be a first for me as it's been awhile since I messed with any of this, I don't really remember.
>
> It seems more likely to me that the problem is either in the Read or the file from the IFS or it's in the send across the socket.
>
> Thoughts, opinions suggestions and anything else is welcome.
>
> thanks,
>
> Shannon O'Donnell