very odd problem...

Anon

Hi all,

I have a really odd problem here.

Here is the scenario. I am storing perl scripts in a MySQL text field - these are pulled and put in a file, which I chown to our client and move to their respective cgi-bin when they want to install it.

All goes great, the file is owned by the client, it is mode 755, everything is peachy - but the script doesn't work.

I ran diff on the script next to a working script-- and it says 'no new line at the bottom of file' - Okay, so I hit a hard return at the end of the file (using pico to test) - and run the script again - works great now. However, that wasn't the problem.

I tried putting a hard return when the script is installed from MySQL - and all the rest of that jazz - nothing seems to effect this error.

If I open the file with pico... add a # char anywhere at all in the script which doesn't effect it's workings, and save and quit, the script will start working. Even if I add and remove the #, as long as I save the file via pico, the script starts working.

Isn't that really odd? Any ideas at all why this is happening.

Just to add to this : Apache log for this virtual host reports 'premature end of script headers' - which really doesn't seem accurate.

I appreciate you reading this really long post.

-m.

deckrdx

How are you creating the file?

Anon

Chad,

the file is created with

exec("touch $gen_filename");

if ($genFILE=fopen("$gen_filename", "w"))
{
// File opened.
fputs($genFILE,$s_info_code,strlen($s_info_code));
}

Thank you for your response.

Regards,
-m.

Anon

Just to add to this - I read in another forum that 'a simple edit via pico' will fix the common error of 'premature script headers'. (which just confirms what I originally posted.)

The reason listed was : the information at some stage is passed as 'binary'. (however fopen documentation lists it's not relevant on unix systems- so.. more confusion).

I don't want to get into opening a pico based pipe to the file - as it's really not the best solution.

(Also, I don't know how to send 'Control+X' to stdin - does anyone?) 🙂

Again, thanks for your help.

Regards,
-m.

deckrdx

I did a bit of searching, but nothing concrete. A couple of the questions I have is...

Are you closing the file pointer with fclose(), or just letting php close it on script end? Maybe closing it with fclose may complete the file better.

Does the script contain the 'print "Content-type: text/html\n\n";', and how is it handled when there are carraige returns in the string you are putting? This is more or less grabbing in the air, but it might be something. Considering it only takes and edit and save of pico to fix, this is probably not the issue.

A popular issue on the web is the permissions the file is set to and what permission is the webserver trying to access it as. Again, with the pico edit, this is probably not the case, but its something to check out. When you fputs the file, change the permissions to 755, or 777 (for testing only), to see if that clears anything up.

Is there an error message before the 'premature end..' message and what is it?

Maybe make sure the "Content-Type" header is printed in the beginning of the script so that if the script is erroring out, you will get the message in the browser.

Does the script run from the command line before editing with pico?

How was the file placed in the database? If it was from a windows system, it might have the \r\n in there, and when you run pico, it converts the \r\n to just \n. Worse yet, if it originated from a mac, it would just have the \r.

I know none of these are THE answer, but hopefully they will spark something for you to find the problem.

Anon

Chad,

firstly, thank you very much for taking the time to respond in such detail.

About the script, it's stored in MySQL on a linux server.

The file is being generated on a secure port which runs with sudo to execute commands not permitted by anyone but root. (Had to go that route for user manager options and qmail functions).

The operation proceeds in the following way :

Script code is called from MySQL
Script is written to file
File permission is changed to 755
File ownership is given to client
File is moved to client's cgi-bin

When I check the file : all operations are successful. The mode is 755, the owner is the client.

About specific questions you had :

Does the script contain the 'print "Content-type: text/html\n\n";

Yes. It's actually the popular FormMail.pl script written by Matt Wright. (http://www.worldwidemart.com/scripts/)

Is there an error message before the 'premature end..' message and what is it?

Only error message :
[error] [client IP_ADDR] Premature end of script headers: /path/to/perl_file.pl

Maybe make sure the "Content-Type" header is printed in the beginning of the script so that if the script is erroring out, you will get the message in the browser.

I do not get this message in the browser.

The source for the 500 error is :

<!DOCTYPE HTML PUBLIC "-//IETF//DTD HTML 2.0//EN">
<HEAD>
<TITLE>500 Internal Server Error</TITLE>
</HEAD><BODY>
<H1>Internal Server Error</H1>
The server encountered an internal error or
misconfiguration and was unable to complete
your request.<P>
Please contact the server administrator,
admin@address and inform them of the time the error occurred,
and anything you might have done that may have
caused the error.<P>
More information about this error may be available
in the server error log.<P>
<HR>
<ADDRESS>Apache/1.3.17 Server at vhostdomain.com Port 80</ADDRESS>

</BODY>

Also,

perl -c script.pl says Syntax OK
perl -e script.pl says nothing

How was the file placed in the database?

PhpMyAdmin - pasted it into the text field.

Right now I am moving on to looking for a way to execute pico and make a small change (control+k, control+u) to the file and saving/exiting out. Not the best way - but maybe it'll work.

Unfortunately I don't know how to send Control+X via popen, but that's a small issue right 🙂

Again, thank you for your response.

I'll keep digging, and if I come across a fix, I'll be sure to share it here.

Best Regards,
-m.

Anon

Just wanted to update all:

I have not found a direct resolution to this problem, however, I tried the following command sequence after the generated file was made ready - and it fixes the problem :

// open pipe to generated file.
$fpop=popen("pico $gen_filename","w");

// Delete the first line read
fputs($fpop,chr(27));
fputs($fpop,chr(27));
fputs($fpop,"k");

// Undo the delete
fputs($fpop,chr(27));
fputs($fpop,chr(27));
fputs($fpop,"u");

// Send file quit command
fputs($fpop,chr(27));
fputs($fpop,chr(27));
fputs($fpop,"x");

// Yes to save
fputs($fpop,"y");

// Enter
fputs($fpop,"\r");

// Close pipe.
pclose($fpop);

Regards,
-m.

Anon

pico automatically appends a \n (newline) to the end of every file it saves. if this makes your scripts work, just add a \n to the file when you create it in your php script.

-paul
http://xhawk.net

Anon

Paul,

One of the first things I tried was \n.

I put a \n in the script itself in mysql - appended one with popen using vi too.

It's basically a 'no change' required pico save which fixes it.

Odd.

Regards,
-m.

deckrdx

Only other thing I can think of is to place the script on the servers filesystem and do a copy instead of taking it from the database. What is storing the script in the database gaining you?

Anon

Not your original question, but just in the interest of decreasing the load on your webserver--why do you have:

  exec("touch $gen_filename");

preceding the creation of the file in:

  if ($genFILE=fopen("$gen_filename", "w"))

Surely the fopen(...,"w") will create the file just fine all by itself, saving the work of loading and running touch just to create an empty file.

Anon

How sure are you that the file is being created/copied with the sequence \r\n ending the header lines, and not just \n?

Anon

But again, it's not the lack of a \n at the end of the script that's likely to be a problem. The error message says "premature end of script HEADERS", not "premature end of script". Again, check that the header lines (and the blank separator line between the headers and the content) are terminated with \r\n and not bare \n!

Anon

I actually didn't know "w" would create the file too. I removed it yesterday after Chad's initial message (as it inclined me to look into the fopen function).

Regards,
-m.

Anon

Chad,

I thought about that. Simply have the web server copy the file.

One thing though, MySQL stores the script, the filename, the function name which is called to update the script for the client's domain (referrer field) - since the software library for clients will have about 100 scripts, I wanted to store them all in a database with or without function calls based on the script.

Again, that doesn't mean the script code itself cannot be a local file.

However, with multiple clients try to log in and issue the copy command on the same instance - will that present a problem? - I just feel storing it in the database is a clever way of doing it, and I already went through hell with file locking and qmail!

I did seriously think about storing the files on the server, but with pico solving the problem, I don't want to go that route.

Regards,
-m.