Problem executing a selfmade linux executable from within PB

AND51 · Post by **AND51** » Tue Apr 28, 2009 2:53 am

Hello!

I almost finished my "PB Interpreter", which allows me to write my whole websites in PureBasic.

Unfortunately, I got a strange problem: My interpreter (a console app) cannot run another console app.

I've got this console executable to be compiled and executed:

Code: Select all

OpenConsole()
PrintN("Content-Type: text/html")
PrintN("")
Print("Hello from PureBasic "+Str(Random(999)))

The code gets compiled correctly, the executable is being created in /tmp/. I can even run it in the terminal and I get this output:

output wrote:Content-Type: text/html

Hello from PureBasic 561

But when my interpreter tries to run the executable, an error occurs. I use this code:

Code: Select all

OpenConsole()

; other code

Define program=RunProgram(exe$, commandline$, "", #PB_Program_Open|#PB_Program_Read|PB_Program_Write)

Define *output, available

While ProgramRunning(program)
    available=AvailableProgramOutput(program)
    If available
        *output=AllocateMemory(available)
        If *output
            ReadProgramData(program, *output, available)
            ; pass the output back to Apache by writing it to OUR stdout
            WriteConsoleData(*output, available)
        EndIf
    EndIf
    Delay(10)
Wend

I cannot read the output and STDERR only contains "Speicherzugriffsfehler (Core Dumped)" (english: Memory Access Error).

What is going wrong here? I'm new to Linux/Unix, but I already got some basic knowledge, i. e. about pipes, stdin, stdout, stderr, bash/shell, etc.
I can also provide this executable so that you can download and run it on your machine for testing.

Oh yes, and my server is running with PB 4.30 and openSUSE 10.3.

bembulak · Post by **bembulak** » Tue Apr 28, 2009 10:44 am

[OT:]

I almost finished my "PB Interpreter",

Sounds interesting! Is is possible to use it as an interactive Shell, like in Ruby or Python?
This would be great!

AND51 · Post by **AND51** » Tue Apr 28, 2009 5:02 pm

Well, I don't know Ruby or Phyton in particular.

When I log into my server via SSH, I can type "/purebasic/interpreter /purebasic/test.pb". Then my interpreter takes test.pb, compiles it with the PB compiler, reads the outout of test.exe and passes this output back to the terminal.

Additionally, my interpreter can also be executed by the webserver Apache. So, if you call www.and51.de/cgi-bin/guestbook.pb, you will get a website generated by PureBasic code and not by Perl or PHP. Is this what you mean?

However, I currently encounter problems with executuing a freshly compiled executable. Neither can I run it, nor can I read its output.
I plan to publish my project, when it's finished, so that you can also use PureBasic for your website on your virtual server.
But this may take a while, as I am very new to Linux/Unix. There are so many tiny things that bug me, e. g. different user accounts, access only via SSH, working via command-line everywhere, etc...
I might finish faster, if you are able to help me, if you want. I am grateful for every kind of support.

walker · Post by **walker** » Tue Apr 28, 2009 6:32 pm

may you have to put an & at the end of your 2nd program (you call within your console-app) to get it running independently from your pgm... just a guess

AND51 · Post by **AND51** » Tue Apr 28, 2009 7:14 pm

Hello walker,

Thank your very much for your advice. I immediately tried to run it with a trailing apmersand, but it doesn't seem to work either.
When I try to append a & when I run my executable from the console, the output does not appear, because it is executed in the background.

Now I continued working and tried it again and again, with and without ampersand, and I achieved a success:
Currently it works, I can run it without ampersand.

Let's see, how long I'm lucky...

AND51 · Post by **AND51** » Tue Apr 28, 2009 8:01 pm

Well, I got the next problem, but this one is not so big as the previous one.

A compiled executable can now be executed and read. But it won't end?

This executable

Code: Select all

OpenConsole()
PrintN("Content-Type: text/html")
PrintN("")
PrintN("<pre>Hallo von PureBasic"+Str(Random(999)))
PrintN("Aktuelles Verzeichnis: "+getcurrentdirectory())
PrintN("ProgramFilename: "+ProgramFileName())

Print("EXIT")
CloseConsole()
End

is being executed and read by this piece of code:

Code: Select all

OpenConsole()
; other code

Define program=RunProgram(exe$, commandline$, dir$, #PB_Program_Open|#PB_Program_Read|#PB_Program_Write|#PB_Program_Error)

Define available, *output
While ProgramRunning(program)
	available=AvailableProgramOutput(program)
	If available
		*output=AllocateMemory(available)
		If *output
			ReadProgramData(program, *output, available)
			; redirecting programs output directly to apache by writing it to OUR console (our stdout)
			WriteConsoleData(*output, available)
		EndIf
	EndIf
	Delay(10)
Wend


PrintN("This line will never be printed")
CloseConsole()
End

Don't worry, 'exe$', 'commandline$' and 'dir$' are correctly set.
But why doesn't this loop break, when the sub-program prints "EXIT" and quits itself?

walker · Post by **walker** » Tue Apr 28, 2009 8:49 pm

... two thoughts.

first: a bug in closeconsole() (regarding to the doc, you have to use killprogramm() and closeprogramm() for completely removing a prog started with RunProgram)
I think, closeconsole only uses "KillProgramm" for itself but not closeprogram for disconnecting from it's caller... so the program-id is still valid :roll:

second: A BUG

AND51 · Post by **AND51** » Tue Apr 28, 2009 9:22 pm

Well, at the moment, I don't have a clue. I'm confused:

1) Using CloseProgram() does not only close STDIN, but also STDOUT. So when I finished my input to that sub-executable and I use CloseProgram() then, I cannot read its output.

2) The sub-executable is correct. So the mistake must be somewhere else. But is my interpreter really wrong? Do I have made a logical mistake? I don't think so.

3) I can't the the sub-executable in the taskmanager (top). So I assume it quits correctly. // Edit: It must be a bug: I can get the ProgramExitCode(), but it is still running. So either the bug is in ProgramRunning() or, as you said, in CloseConsole().

I will post this in the bug section.

walker · Post by **walker** » Tue Apr 28, 2009 9:37 pm

.. may you can work with a woraround

simply send a special string (something unique) just before quitting the called app and if it's read by your programm, use kill/Closeprogramm to cut off the connection ... may that's a way untill the bug (still considering it is one) fixed

AND51 · Post by **AND51** » Tue Apr 28, 2009 9:58 pm

Yes, I already thought of this solution, too.
If there won't be quick help, I must do so until the bugfix.

Meanwhile, I posted this in the Linux bugsection.
There is a link to the script that is compiled and executed everytime, you visit the page. So you can see my interpreter live!

@ bembulak:
You may click this link, too. I'm sure, you will, as you are interested.

Note: there is also some debugging output, so don't wonder.

walker · Post by **walker** » Wed Apr 29, 2009 12:32 am

hey..... that's pretty cool

(tried your link)

you can prog an app using your webbrowser simply by sending a .pb source

AND51 · Post by **AND51** » Wed Apr 29, 2009 6:27 am

Yes, you are right. First, I wanted only my website to be in PureBasic, but then I got aware of the fact that I can also open a port on the server side to receive .pb-files. I'm sure there are some possibilities.

AND51 · Post by **AND51** » Wed Apr 29, 2009 10:06 pm

Hey walker and bembulak (and all others),

I published this project in the german forum: http://www.purebasic.fr/german/viewtopi ... nterpreter

You can participiate there. I will publish it in this forum, too, when I collected some ideas from the german members.