World: r3wp

It might be faster too.

Had the same result ...

Since my encapped applications work fine under Firedaemon, I am going 
to try and encapCheyenne next.

This was all I was doing:


call/wait "gswin32.exe -q -dNOPAUSE -dBATCH -sDEVICE=pdfwrite -sOutputFile=test.pdf 
-f test.ps"

https://www.compkarori.co.nz/cgi-bin/showfile.rsp?key=#{E7F8390D60C36AFFB11252677A3586A5}

gives an empty value for key

remove the # and it's okay.

bug report :)

it's bug in Rebol's decode-url:

>> probe decode-url https://www.compkarori.co.nz/cgi-bin/showfile.rsp?key=#blabla
make object! [
    user: none
    pass: none
    host: "www.compkarori.co.nz"
    port-id: none
    path: "cgi-bin/"
    target: "showfile.rsp?key="
]

Interesting .. should we Rambo it?

Or, is it because # is not a legal character in a url?

Is there a limit to how much you can send in a Post command?

As you might have read from above, I am attempting to post postscript 
to a RSP script to convert to PDF.
4k of postscript sees the script complete.
But 92Kb of postscript - and I get this response

make object! [
    code: 800
    type: 'user
    id: 'message

    arg1: {Error.  Target url: https://www.compkarori.co.nz:443/cgi-bin/createpdf.rsp 
    could not be retrieved.  Server response: none}
    arg2: none
    arg3: none

    near: [page: read/custom https://www.compkarori.co.nz/cgi-bin/createpdf.rsp
    reduce ['POST join "content=" content]]
    where: none
]

I think a POST should be okay with 100kb of text .. so perhaps there 
is problem with Cheyenne's handling?

Maybe I should try and simulate a multipart form post on the client 
side instead??

Check your httpd.cfg file to see if you find a 'post-max directive.

If you find one, set it to higher limit (unit is Kb) or just remove 
it.

G: # is used to link to anchors.  e.g. http://www.rebol.com/cgi-bin/blog.r?view=0338#comments

you need to encode # with percent encoding, and you need to make 
sure rebol is nod decoding that for you silently (the url! datatype 
does this unfortunately).

# is a legal url char... Rebol's decode-url is full of bugs for example 
this url is bad decoded as well: http://test/path/target?text/something

I did a grep on the sources,and I can only find post-max in 2 files 
...

mod-static
post-max: [integer!] in main

httpd.r
limit: select req/cfg 'post-max

Not in httpd.cfg

I guess I'll just create a limit in httpd.r

Hmm.  Didn't help either.  Still same error.

Checking the server logs .. looks like the post is not even registering 
there either.

does it work if you test it locally without SSL ?

oldes, it is legal but NOT part of query args UNLESS percent-encoded. 
special characters in uris must always be percent-encoded when not 
used as separators.

Good question .. yes, it works remotely without ssl.

Cheyenne is running on port 8100, and SSL is directed to that from 
stunnel.

So, if you try 


page: read/custom https://www.compkarori.co.nz/show.cgireduce [ 
'post join "content=" content ]  
then this fails if content is a few kbs

but if you do 


page: read/custom http://www.compkarori.co.nz:8100/show.cgi reduce 
[ 'post join "content=" content ]

it works fine.

What a shame.

So it seems that the issue is either with the HTTPS protocol in REBOL 
or with stunnel.

I guess so

So, I need to mock up a webpage to do the same post

and check it out in a browser

Exactly.

Hmm.

I don't have a https server I can test ...

Be sure to use the latest Cheyenne beta (0.9.15). It fixes a regression 
bug on multipart data decoding.

Do you have a log of which files are changed?

Or do I have to update every file?

I guess I'd better download curl and use that to test with

Sorry, no diff file list. You can update just the files having a 
more recent timestamp.

Ok

curl seems to work

curl -k -F upload=@test.ps -F press=OK https://www.compkarori.co.nz/show.cgi

So, I guess this means it's a Rebol https problem :(

Wasn't Gabriele asking for people to test a new http(s) protocol 
??

Does anyone wish to independently confirm my findings with https?

Rebol []

data: copy/part read %test.ps 1024
content: copy data
for i 1 100 1 [
	cnt: i * 1024

 page: read/custom https://www.compkarori.co.nz/cgi-bin/measure.rsp
 reduce  [ 'post join "content=" content ]
	parse page [ thru <data> copy received to </data> ]
	either found? received [
		received: to integer! trim received 
		either received = cnt [
			print [ "OK at " cnt ]
			append content data
		][
			print [ "oops .. failed at " cnt ]
			break
		]
	][
		print "failed to read ... "
		halt
	]
]

http continues okay.
https fails at 15360 bytes