need help, will pay

[boman]

Need help with a simple Perl cgi script running from Apache 2.2 on
Windows XP, willing to pay $100 for the solution. Please email me for
details. Prefer NYC-based coder, payment in cash.

 

[Nathan Keel]

Need help with a simple Perl cgi script running from Apache 2.2 on
Windows XP, willing to pay $100 for the solution. Please email me for
details. Prefer NYC-based coder, payment in cash.

Why don't you post what the error is and the relevant portions of the
script? You can probably get good help for no charge right now.

 

[boman]

Why don't you post what the error is and the relevant portions of the
script?  You can probably get good help for no charge right now.

This question has been asked before, and I've tried almost all of the
solutions, with no success. So I'm trying not to annoy this group by
asking the question again. After the reading the "why doesn't this
work" thread from Aug 20, I didn't want to start another brawl.

But you graciously asked, Nathan, so here goes!

1. display web page with simple form w/ radio buttons
2. user selects radio button and clicks submit
3. single parameter is passed from the radio button
4. exe command on web server is initiated (I was using
Win32:rocess), parameter is passed
5. output of command is sent back to browser in real time (this is
where I'm stuck)
6. when command is done, start 2nd command, send output to browser in
real time

I've got 1-4 working, but cannot get output of command to the browser
in real time, as the output is generated. It simply waits for command
to finish, then sends the entire out put in one blob.

Yes, I've got the $|=1 set, and I made sure apache is not buffering
output, but can't get around the browsers doing the buffering (I think
this is what's happening).

Yes, I've read the stonehenge article which uses traceroute as an
example, but cannot get that to work on my Windows perl v5.8.8. I
tried some of the concepts from this article, but it's a bit out of my
league.

At this point I cannot spend much more time trying to get this to work
myself, and I need a simple, yet professional level solution. I've
read lots of forums, paged thru numerous books, I can't really spend
time on that anymore. I need something that works, and I need it this
week.

I know this can probably be done in javascript, but I'd prefer this to
be done in Perl. Also, I won't be able to install any modules, we'll
have to use what's installed by default.

thanks!

 

[Brian Wakem]

This question has been asked before, and I've tried almost all of the
solutions, with no success. So I'm trying not to annoy this group by
asking the question again. After the reading the "why doesn't this
work" thread from Aug 20, I didn't want to start another brawl.

But you graciously asked, Nathan, so here goes!

1. display web page with simple form w/ radio buttons
2. user selects radio button and clicks submit
3. single parameter is passed from the radio button
4. exe command on web server is initiated (I was using
Win32:rocess), parameter is passed
5. output of command is sent back to browser in real time (this is
where I'm stuck)
6. when command is done, start 2nd command, send output to browser in
real time

I've got 1-4 working, but cannot get output of command to the browser
in real time, as the output is generated. It simply waits for command
to finish, then sends the entire out put in one blob.

Yes, I've got the $|=1 set, and I made sure apache is not buffering
output, but can't get around the browsers doing the buffering

A quick and easy way around browser buffering is to print something like
this 100 or so times after every line you output.

<!-- this wont show up on the page but will force the browser to flush -->

 

[Jürgen Exner]

1. display web page with simple form w/ radio buttons
2. user selects radio button and clicks submit
3. single parameter is passed from the radio button
4. exe command on web server is initiated (I was using
Win32:rocess), parameter is passed
5. output of command is sent back to browser in real time (this is
where I'm stuck)
6. when command is done, start 2nd command, send output to browser in
real time

I've got 1-4 working, but cannot get output of command to the browser
in real time, as the output is generated. It simply waits for command
to finish, then sends the entire out put in one blob.

As has been explained to you back then this is _NOT(!)_ a Perl issue but
a problem somewhere in the long chain of receiving CGI response,
creating and sending HTTP response, receiving HTTP response, rendering
the page in the browser. You really need someone who is familiar with
those areas (no matter in which language) instead of someone who is
familiar with Perl.

In short: you are barking up the wrong tree.

Yes, I've got the $|=1 set, and I made sure apache is not buffering
output, but can't get around the browsers doing the buffering (I think
this is what's happening).

As I said, you need someone who is familiar with web programming, not
someone familiar with Perl.

I know this can probably be done in javascript, but I'd prefer this to
be done in Perl.

That is not possible! Please, please, please, you already set the Perl
buffer to auto-flush. There REALLY is nothing else you can do in your
Perl code.

You said yourself that the buffering probably happens in the brower.
Maybe so, maybe somewhere else in the chain. But it really, really isn't
Perl.

Aside of that, "in real time" is a non-starter to begin with because the
HyperText Transfer Protocol is not real-time to begin with.
If you need real time then you need to look into a different protocol,
but that's even more off topic than web programming.

jue

 

[boman]

As I said, you need someone who is familiar with web programming, not
someone familiar with Perl.

You said yourself that the buffering probably happens in the brower.
Maybe so, maybe somewhere else in the chain. But it really, really isn't
Perl.

Jürgen, I've read your previous posts and I respect your opinion, but
I'm pretty sure what I'm asking for can be done in Perl, the
stonehenge article with the traceroute example seems to prove that. I
simply cannot spent any more time trying different things.

Hence the offer of money; if someone can provide me with a solution
based on the requirements (which include using Perl), than I will pay
cash.

Aside of that, "in real time" is a non-starter to begin with because the
HyperText Transfer Protocol is not real-time to begin with.
If you need real time then you need to look into a different protocol,
but that's even more off topic than web programming.

OK, not real time, how about updating the browser every 5 seconds? I
actually do have something like this working, but I can't get around
the browser buffer issue, and I'm pretty sure someone who knows Perl
very well can provide a clever solution to get around that. I simply
cannot spend any more time researching this, I need a solution now.

thanks

 

[Nathan Keel]

Yes, I've got the $|=1 set, and I made sure apache is not buffering
output, but can't get around the browsers doing the buffering (I think
this is what's happening).

I see, well then you might need to get someone to take a look or give
you different options. There are some things you can try, but if
you've set the buffering in the script, then the problem does lie
elsewhere. You also might consider something like a server-push
solution or just saying "click here to view the results", which can
then open a file or read in the data as its happening (perhaps not via
a CGI script).

 

[boman]

I see, well then you might need to get someone to take a look or give
you different options.  There are some things you can try, but if
you've set the buffering in the script, then the problem does lie
elsewhere.  You also might consider something like a server-push
solution or just saying "click here to view the results", which can
then open a file or read in the data as its happening (perhaps not via
a CGI script).

The link idea could be a fall back, but I'd like to automate that. I
would think you can do that by setting the refresh parameter in
cgi:m's header call, but I cannot figure out how to get the refresh
to exit once the process is complete, it just refreshes and refreshes
and refreshes, even when the process completes. The stonehenge
traceroute script gave me ideas, but I can't get it to work, and I'm
out of time for troubleshooting.

 

[Jürgen Exner]

Jürgen, I've read your previous posts and I respect your opinion, but
I'm pretty sure what I'm asking for can be done in Perl, the
stonehenge article with the traceroute example seems to prove that. I
simply cannot spent any more time trying different things.

Ok, one final attempt:

You are trying to haul a yard of sand with your pickup truck and it
doesn't work. You keep trying to convince the car mechanic to finally
fix the car while actually the problem is the washed out bridge to the
sand pit.

This is exactly what you keep doing. You are looking for a solution in
Perl (your pickup) and keep asking Perl guys (your car mechanic) how to
fix the program (your pickup) while actually Perl (your pickup) is doing
just fine and the problem is not related to Perl (your pickup) at all
and you should be asking web programmers (the DOT) about how to
implement a non-bufferd path (fix the bridge).

OK, not real time, how about updating the browser every 5 seconds? I

If you mean updating the displayed document rather than the browser,
then HTML has such a directive, see "refresh" for HTTP-EQUIV. If that's
a good choice I don't know, probably not.

jue

 

[Jürgen Exner]

The link idea could be a fall back, but I'd like to automate that. I
would think you can do that by setting the refresh parameter in
cgi:m's header call, but I cannot figure out how to get the refresh
to exit once the process is complete, it just refreshes and refreshes
and refreshes, even when the process completes.

That is trivial to fix: for the final report, i.e. when the process is
finished, just _don't_ set the refresh parameter in that final document.
Problem solved.

jue

 

[Jens Thoms Toerring]

The link idea could be a fall back, but I'd like to automate that. I
would think you can do that by setting the refresh parameter in
cgi:m's header call, but I cannot figure out how to get the refresh
to exit once the process is complete, it just refreshes and refreshes
and refreshes, even when the process completes.

That should be simple, just don't send the meta refresh anymore
with the last page when the program is done.

The real problem is more that this can't be done with a simple
CGI script because when the browser (re-)loads the page the
web server will start a new CGI script that doesn't know any-
thing about what happened before. So you will probably need
some background process that waits for requests from the CGI
script and then passes back the newly calculated data back to
it (which the CGI script then sends on to the web server that
transmits them to the browser).

Since there could be several browsers asking simultaneously
for new pages you also would have to be able to handle cal-
culations in parallel within that background process. And
you would have to rely on some mechanism, e.g. cookies, to
know which results are for which client.

Regards, Jens

 

[boman]

Here's the code so far. It continues to update the browser, although
it's sloppy, and when the process is done, it does display the "done"
page. But then it refreshes again in 5 seconds.



#!c:/Perl/bin/perl.exe

use CGI qw/:standard/;
use Win32:rocess;
use Win32:rocess qw/STILL_ACTIVE/;
use Win32;

$| = 1;

Win32:rocess::Create($MainProcessObj,
"c:\\WINDOWS\\System32\\netstat.exe",
"netstat",
0,
NORMAL_PRIORITY_CLASS,
".");
$MainProcessObj->GetExitCode($exitcode);

$counter = 0;
while ($exitcode != 1) {
if ($exitcode == 0) {
print header, start_html,h4("something is wrong"), end_html;
exit;
} else {
print header(-Refresh=>"5"), start_html;
print p("waiting, $counter secs... $exitcode "), end_html;
$counter = $counter + 5;
sleep 5;
$MainProcessObj->GetExitCode($exitcode);
}
}

if ($exitcode = "") {
print header, start_html, p("done $exitcode"), end_html;
}

 

[Nathan Keel]

Here's the code so far. It continues to update the browser, although
it's sloppy, and when the process is done, it does display the "done"
page. But then it refreshes again in 5 seconds.

If you're happy with the result as it is and just don't want it to
refresh, just pass a variable to the script that let's the script know
that it's been refreshed (once) already and doesn't put in the refresh
header in that resulting document's (web page) output.

 

[Ron Bergin]

Here's the code so far. It continues to update the browser, although
it's sloppy, and when the process is done, it does display the "done"
page. But then it refreshes again in 5 seconds.

#!c:/Perl/bin/perl.exe

use CGI qw/:standard/;
use Win32:rocess;
use Win32:rocess qw/STILL_ACTIVE/;
use Win32;

$| = 1;

Win32:rocess::Create($MainProcessObj,
        "c:\\WINDOWS\\System32\\netstat.exe",
        "netstat",
        0,
        NORMAL_PRIORITY_CLASS,
        ".");
$MainProcessObj->GetExitCode($exitcode);

$counter = 0;
while ($exitcode != 1) {
        if ($exitcode == 0) {
                print header, start_html,h4("something iswrong"), end_html;
                exit;
        } else {
        print header(-Refresh=>"5"), start_html;
        print p("waiting, $counter secs... $exitcode "), end_html;
        $counter = $counter + 5;
        sleep 5;
        $MainProcessObj->GetExitCode($exitcode);
        }

}

if ($exitcode = "") {
        print header, start_html, p("done $exitcode"), end_html;

}

Sloppy is an understatement.

Why are you printing/reprinting the html header in a loop and why the
'sleep'?

You should execute that script outside of the cgi environment and
remove the 'sleep' line to see why that's the wrong thing to do.

Rather than using Win32:rocess, I think it would be better to use
Net::Netstat::Wrapper.

http://search.cpan.org/~mcantoni/Net-Netstat-Wrapper-0.03/lib/Net/Netstat/Wrapper.pm

 

[Martijn Lievaart]

Yes, I've got the $|=1 set, and I made sure apache is not buffering
output, but can't get around the browsers doing the buffering (I think
this is what's happening).

Some (all?) versions of IE need 256 spaces first as start of the body.
It's weird, but true. If your problem occurs in other browsers as well,
this is not the solution, if your problem is only in IE, give it a try.

HTH,
M4

 

[Ron Bergin]

Yes, I've got the $|=1 set, and I made sure apache is not buffering
output, but can't get around the browsers doing the buffering (I think
this is what's happening).

If you run the script from the command line, you'll see that the issue
not that the browser is buffering the netstat output, the netstat
command is buffering its output.

 

[Peter J. Holzer]

That should be simple, just don't send the meta refresh anymore
with the last page when the program is done.

The real problem is more that this can't be done with a simple
CGI script because when the browser (re-)loads the page the
web server will start a new CGI script that doesn't know any-
thing about what happened before.

I think you haven't read the "stonehenge article" mentioned by boman.
Here it is: http://www.stonehenge.com/merlyn/LinuxMag/col39.html

So you will probably need some background process that waits for
requests from the CGI script and then passes back the newly calculated
data back to it (which the CGI script then sends on to the web server
that transmits them to the browser).

No, the idea is to have a long-running process just write information to
a file. The CGI script then just has to look at the file to determine
what it should send back to the client.

Since there could be several browsers asking simultaneously
for new pages you also would have to be able to handle cal-
culations in parallel within that background process.

Just use several background processes.

And you would have to rely on some mechanism, e.g. cookies, to know
which results are for which client.

Query parameters are more reliable for that.

hp

 

[Jens Thoms Toerring]

I think you haven't read the "stonehenge article" mentioned by boman.
Here it is: http://www.stonehenge.com/merlyn/LinuxMag/col39.html

No, I hadn't since I could find it. Thanks for the link.

Regards, Jens