World: r3wp

When it works you get :

NEW/LOCALS is an object of value:

   handler         object!   [name port-id hidden peer stop-at shared 
   module ev...
   write-queue     block!    length: 0
   file-chunk      integer!  65536
   stop            none!     none
   in-buffer       binary!   #{}
   file            none!     none
   flag-close      none!     none
   events          none!     none

When it crash I have :

NEW/LOCALS is an object of value:
   write-queue     block!    length: 0
   file-chunk      integer!  65536
   stop            char!     #"^/"
   in-buffer       binary!   #{}
   file            none!     none
   flag-close      none!     none
   events          none!     none

** Script Error: Invalid path value: handler
** Where: init-connection
** Near: list: array len: length? names:

The problem is simple : new/locals is set at the beginning of the 
function and it's content has changed 3 lines later !!

It's 100% reproducable, but with precise conditions. It's called 
by a timer (manually, I haven't get any error).

With a timer of 2 mins, it happens the second time. With different 
time it does not happen, or I haven't wait enough (but did not make 
to much test on that)..

I don't know how I can post that to Rambo !

Sorry to flood this group.

could it be that reduce user-data is changing something?

Yes, please confirm user-data is not doing anything.  Then, add an 
empty word/value at top of locals object.  Does that same cell get 
deleted?

What makes this bug very unusual is that the rest of the object seems 
to be ok. It is as if something deleted the word from the object 
(but that is never done in REBOL).

Gabriele: #3402 is an outstanding bug when using /CUSTOM refinement. 
 An example is submitting a  form over http using REBOL. You need 
to specify both HEADER and POST. For example:

   read/custom url [HEADER [Content-Type: {multipart/form-data; boundary=---xyz} 
   ] POST {blah}]

The problem is that the handling of POST *overwrites* the Content-Type 
and Referer even though they may have been specified in HEADER. This 
logic bug is within the create-request function of Open in HTTP scheme.

Gabriele: #3016 still applies but is less likely to occur with the 
increased limits.  On View 1.2.46.3.1 my test failed after 100 To 
addresses, now on 1.2.107.3.1 my test fails after 256 To addresses. 
My test is next:

test-address-import: func [

 {Returns true if pass, false if discrepency and none if failed with 
 error.}
	limit [integer!] "Number of TO address to generate."
	/quiet "Does not display error."
	/local to-list eml sep msg obj
] [
	to-list: copy {} sep: ""
	repeat i limit [
		eml: to-email join "test-" [i "@test.com"]
		repend to-list [sep {"'} eml {'" <} eml {>}]
		if empty? sep [sep: {,^/	}]
	]

	msg: replace copy {Date: Thu, 12 Feb 2004 11:41:49 +0100
From: test <[test-:-test-:-com]>
MIME-Version: 1.0
To: TO_LIST
Subject: [REBOL] test message - edited copy of real message
Content-type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 8bit
Sender: [rebol-bounce-:-rebol-:-com]
Reply-to: [testing-:-testing-:-com]
Status:   


Test message.
} {TO_LIST} to-list

	either error? set/any 'result try [
		obj: import-email msg
		limit = length? obj/to
	][if not quiet [print mold disarm result] none][result]

]

This test generates a test email with the desired number of addresses 
then attempt to import it.

Romano did a lot of work to try to overcome this by trying to take 
recursion out of the code.

#3402: i see what you mean, but you should use:

read/custom url [POST {blah} [Content-Type: {multipart/form-data; 
boundary=---xyz} ]]

which works. however, i'll look into it.

Important: 1.2.108 is out. Please, test it. There are a number of 
mezzanine changes that need testing to ensure we are not breaking 
anything.

1.2.108 supports SMTP Auth, but it has not been tested.

note that this has been added as a new scheme, esmtp://. Now SEND 
uses esmtp:// and not smtp://. We have left smtp:// for compatibility 
reasons.

to set your user name and password, use set-net.

(see help set-net)

note that if you set esmtp-user to the word 'ask, it will ask you 
user name and password (on the console) the first time you send an 
email.

we have changed LAYOUT as suggested in #3341. Please let us know 
if this may create problems.

SPLIT-PATH and TO-IDATE have been changed too as suggested in #3622 
and #3623. Please test them well.

header parsing has been changed to support multiple headers (fixes 
#3090). this MIGHT break some code that does not expect getting a 
block instead of a string in the returned object. feedback is welcome.

other changes include: query on http port uses HEAD instead of GET, 
and fixed inform/timeout. please check out RAMBO to see a list of 
all the other bugs that have been fixed.

http://www.rebol.net/cgi-bin/rambo.r?sort=6&limit=3&cmd=Search&id=&pattern=&subject-only=true

Gabriele, what type of authentication does esmtp do?

AUTH CRAM-MD5 DIGEST-MD5  ?

If only AUTH, would be hard to add CRAM-MD5 which is already supported 
in Rebol's IMAP ?

CRAM-MD5, if it fails it tries PLAIN. i hope we can make it smarter 
in the future by looking into what the server actually supports (EHLO 
command)

Carl, Gabriele : I have created group "Debug p1.3" to continue on 
my bug and for others too

#3700: great! still doesn't work on images, but works fine with binary! 
and string!

some issues with string! usage in 'and 'xor 'or : 

1) string! op string! -> binary!, should returns a string! instead

2) a) string! op binary! -> error!, a) binary! op string! -> error! 
, 
should be allowed and do the same than:

a) string! op as-string binary! -> string! , b) binary! op as-binary 
string! -> binary!

#3697: can't crash REBOL anymore (Win2k) - tried some heavy brush 
pasting on picture: works fine.

#3694: seems fixed: my one-face-draw-textlist no longer needs a workaround 
- pen changes are applied.

#3693: works fine now :)

anyone was able to test smtp auth?

Gabriele smtp-auth is working here. Tested against smtp.mac.com Apple's 
server for their dot-Mac accounts, similar to HotMail.

http://www.rebol.net/notes/rv13-changes.htmlupdated

View 1.2.109 and 1.2.110   no longer have CTX-EDIT, where did that 
code go ?

These enhancements/bug fixes are great! :)

anton: ctx-viewtop/ctx-edit. but, this will be fixed.

#3700: complement fully working, ops issues corrected ^_^

Warning, /View 1.2.11.3.1 and /View 1.2.110.3.1 are both named rebview1211031.exe

Oops, talked a little too fast. Testing all datatypes combinations 
for 'and, 'xor, 'or, 

[ binary! | string! ] op [ logic! | number! | char! | tuple! ]  -> 
crash.

Tested FTP operations against two of my sites that used to require 
%ftp-patch53.r & "system/schemes/ftp/passive: true" to work ... they 
both work fine now with the new FTP protocol handler. Good job Romano 
[and others].

Vincent, that's right, we now have name collisions with older betas. 
Luckily the other day I wrote a script to rename all my rebview executables 
unambiguously. For interest, this is how I extracted the version 
number from each executable: