Mailman 3 March 2000 - Mailman-Developers

French Templates.
by Hausermann Laurent March 15, 2000

March 15, 2000

Hi, I currently finalizing a communuty web server (for a former student association) in French, and I cannot find any ressources about French Templates for mailman 1.1. Do you now about any translation project ? Thanks a lot Laurent Hausermann

1 0

Re: [Mailman-Developers] Failed authentication
by J C Lawrence March 14, 2000

March 14, 2000

On Sun, 12 Mar 2000 20:30:17 -0800 J C Lawrence <claw(a)kanga.nu> wrote: > I have no bloody idea. POSTS to PHP work just fine. POSTs to > Perl CGIs work fine. Once I add the PHP3 module to Apache's > config, POSTs to Mailman stop. > Colour me confused. > As soon as I get few other things resolved, I'm going to see what > happens if I backlevel PHP, or even try PHP4... I've identified the exact setup that causes the behaviour under Debian/potato: If I install the following .debs for PHP everything works fine: php3-gd_3%3a3.0.15-1_i386.deb php3_3%3a3.0.15-1_i386.deb php3-mysql_3%3a3.0.15-1_i386.deb If however I install the current Debian/potato set: php3_3%3a3.0.15-1.2_i386.deb php3-gd_3%3a3.0.15-1.2_i386.deb php3-mysql_3%3a3.0.15-1.2_i386.deb with the same Apache configs in both cases (a LoadModule line and two AddTypes as per the default http.conf and srm.conf), or I install the prior the prior 2.0.15-1.1 package set, HTTP POSTs to MailMan CGIs no longer happen due to the fact that nothing reaches the stdin of the CGIs. It doesn't seem to matter whether I'm running: apache_1.3.9-12_i386.deb apache-common_1.3.9-12_i386.deb or: apache_1.3.9-11_i386.deb apache-common_1.3.9-11_i386.deb or the various libc6 packages in that date range. What I havem't tried, and what I'm going to leave up to Gergely Madarasz (CC'ed on this message) the Debian package maintainer for both the PHP3 and MailMan packages, is finding out if rebuilding everything with a current (or even the same) glibc version makes the problem go away (it hopefully being reproducable at his end). I'm willing to supply copies of my config files if there's interest. I'm going to bed. -- J C Lawrence Home: claw(a)kanga.nu ----------(*) Other: coder(a)kanga.nu --=| A man is as sane as he is dangerous to his environment |=--

3 4

Re: Bogus directory
by Jim Tittsler March 14, 2000

March 14, 2000

On Fri, Mar 03, 2000 at 10:49:45AM -0300, Rodolfo Pilas wrote in mailman-users: [...] > However I am in receipt of a message about bogus directory: [...] > List todos.tecnico has a bogus archive_directory: > /var/spool/mailman/archives/private/todos.tecnico > List todos.sistemas has a bogus archive_directory: > /var/spool/mailman/archives/private/todos.sistemas The "bogus archive directory" issue seems to come up fairly often. Perhaps it would be worthwhile to add a check to cron/nightly_gzip to find out if it is simply a matter of there not yet being any articles to archive. Mailman could check post_id or last_post_time. I've attached one possibility. -- Jim Tittsler, Tokyo Python Starship http://starship.python.net/crew/jwt/

2 1

Re: [Mailman-Developers] Failed authentication
by J C Lawrence March 13, 2000

March 13, 2000

On Sun, 12 Mar 2000 09:50:13 -0800 J C Lawrence <claw(a)kanga.nu> wrote: > The ~mailman/cgi-bin/admin is not getting anything on stdin. Not > good. Current findings: Installation of the PHP3 apache modules under Debian/potato kills POST methods for MailMan. I don't know why. Problem is, I need both. -- J C Lawrence Home: claw(a)kanga.nu ----------(*) Other: coder(a)kanga.nu --=| A man is as sane as he is dangerous to his environment |=--

2 2

a more uniform interface to private and public archives
by Todd Pfaff March 13, 2000

March 13, 2000

for mailman-1.1, the installation documentation suggests to create, for example, an 'Alias' path for the web server to the mailman public archives and a 'ScriptAlias' cgi-bin path to the private archives. what i'd like to suggest is that we make the interface more uniform by eliminating the 'Alias' path and access both private and public archives via a single cgi-bin interface. if the archive is private we require authentication, if not we simply bypass the authentication. i've done this with my mailman installation by doing the following: - created a new version of the MailMan/Cgi/private.py program - in mm_cfg.py, set PUBLIC_ARCHIVE_URL = '/mailman/private' PRIVATE_ARCHIVE_URL = '/mailman/private' - these could then be collapsed into one ARCHIVE_URL - we could also replace PUBLIC_ARCHIVE_FILE_DIR = os.path.join(PREFIX, 'archives/public') PRIVATE_ARCHIVE_FILE_DIR = os.path.join(PREFIX, 'archives/private') with one ARCHIVE_FILE_DIR, and we could also get rid of the public and private subdirectories altogether. in the new private.py i check listobj.archive_private and if it's set to 1 i do the usual private authentication as before. if it's not 1, i set is_auth to 1 and fall through. that's it. very clean and simple. does anyone see any problems with this? i think it certainly makes things more clear and straightforward. -- Todd Pfaff \ Email: pfaff(a)mcmaster.ca Computing and Information Services \ Voice: (905) 525-9140 x22920 ABB 132 \ FAX: (905) 528-3773 McMaster University \ Hamilton, Ontario, Canada L8S 4M1 \

2 1

Re: [Mailman-Developers] Failed authentication
by J C Lawrence March 13, 2000

March 13, 2000

On Sun, 12 Mar 2000 00:19:18 -0800 J C Lawrence <claw(a)kanga.nu> wrote: > Findings: read_urlencoded never reads anything from stdin. Zero. > Zilch. Nada. There's nothing there. The key edits: ... > According to that Apache is sending all the right data, it just > ain't getting there. I've also upgraded libc6 recently (Debian > system FWIW with packages libc6_2.1.3-[567]_i386.deb (notice regex > range)). I downgraded libc6 to the package versions (5 and 6) > that were installed when it did work to no effect. I also rebuilt > the Mailman C based CGI programs under the installed version of > libc6, and copied them over (in case something odd was going on) > to no observed effect. I changed cgi-wrapper.c to read: #include <stdio.h> #include "common.h" /* passed in by configure */ #define SCRIPTNAME SCRIPT #define LOG_IDENT "Mailman cgi-wrapper (" ## SCRIPT ## ")" /* GID that CGI scripts run as. See your Web server's documentation. */ #define LEGAL_PARENT_GID CGI_GID const char* logident = LOG_IDENT; char* script = SCRIPTNAME; const int parentgid = LEGAL_PARENT_GID; int main(int argc, char** argv, char** env) { int status; char* fake_argv[3]; char ac[31]; /* !!!!!!!!!!!edited!!!!!!!!!!!!! */ memset (ac, 0, 31); /* !!!!!!!!!!!edited!!!!!!!!!!!!! */ running_as_cgi = 1; if (getgid()>=100 && getgid()!=65534) check_caller(logident, parentgid)$ /* for these CGI programs, we can ignore argc and argv since they * don't contain anything useful. `script' will always be the driver * program and argv will always just contain the name of the real * script for the driver to import and execute (padded with two dummy * values in argv[0] and argv[1] that are ignored by run_script(). */ fake_argv[0] = NULL; fake_argv[1] = NULL; fake_argv[2] = script; fread (ac, 1, 30, stdin); /* !!!!!!!!!!!!edited!!!!!!!!!!!! */ fprintf (stderr, "---[%s]---",ac); /* !!!!!!!!!!!!edited!!!!!!!!!!!! */ status = run_script("driver", 3, fake_argv, env); fatal(logident, status, "%s", strerror(errno)); return status; } The ~mailman/cgi-bin/admin is not getting anything on stdin. Not good. -- J C Lawrence Home: claw(a)kanga.nu ----------(*) Other: coder(a)kanga.nu --=| A man is as sane as he is dangerous to his environment |=--

2 1

Re: [Mailman-Developers] Failed authentication
by J C Lawrence March 12, 2000

March 12, 2000

On Sun, 12 Mar 2000 15:24:06 +0100 (MET) Gergely Madarasz <gorgo(a)caesar.elte.hu> wrote: > On Sun, 12 Mar 2000, J C Lawrence wrote: >> Background data as this is also going to the Debian Mailman >> package maintainer: >> >> Mailman's web interface suddenly stopped working for me. The >> forms would display but I am unable authenticate not matter what >> browser (Netscape 4.05 and 4.7, Lynx, W3M or Mozilla M14) I use. > Hmm... unable to authenticate ? That means you get to the password > screen each time you try to submit a form ? This usually happens > if you don't access your admin interface on the Base url set for > your list... Nope, that's not the problem (and I'm accessing on the Base Url to boot). The problem is that the Mailman CGIs are not getting the form submission data from Apache. Apache appears to be supplying it (ScriptLog says it is), but by the time the Python code runs its no longer there. -- J C Lawrence Home: claw(a)kanga.nu ----------(*) Other: coder(a)kanga.nu --=| A man is as sane as he is dangerous to his environment |=--

1 0

Re: [Mailman-Developers] Failed authentication
by J C Lawrence March 12, 2000

March 12, 2000

Background data as this is also going to the Debian Mailman package maintainer: Mailman's web interface suddenly stopped working for me. The forms would display but I am unable authenticate not matter what browser (Netscape 4.05 and 4.7, Lynx, W3M or Mozilla M14) I use. On Thu, 09 Mar 2000 13:11:06 -0800 J C Lawrence <claw(a)kanga.nu> wrote: > On Thu, 9 Mar 2000 15:55:44 -0500 (EST) bwarsaw > <bwarsaw(a)cnri.reston.va.us> wrote: >>>>>>> "JCL" == J C Lawrence <claw(a)kanga.nu> writes: JCL> Aye, what's especially odd is that forms are process properly JCL> elsewhere (PHP for instance). So the Apache side and the JCL> modules sides are Okay its just something with the Mailman CGI JCL> wrapper. I guess. >> I still want to know at what point the cgi data is getting >> corrupted. Is it between Apache and the C wrapper? Is it >> between the C wrapper and invoking the Python driver script? Is >> it between the driver script and the specific Mailman/Cgi module? > I'm really not sure, and I'm not too sure how to find out either. > Could you give me some hints where to poke this? I've done some more digging, implemented ScriptLog in Apache, inserted sys.exit(1) in the admin script to force a failure right after the FieldStorage call, and then instrumented pythonlib/cgi.py. Findings: read_urlencoded never reads anything from stdin. Zero. Zilch. Nada. There's nothing there. The key edits: --<cut>-- def read_urlencoded(self): """Internal: read data in query string format.""" qs = self.fp.read(self.length) qs=str(qs) sys.stderr.write (qs) sys.stderr.write ('=======') sys.stderr.write (str(self.length)) ...etc --<cut>-- --<cut>-- %% [Sat Mar 11 23:58:57 2000] POST /lists/admin/library/ HTTP/1.0 %% 500 /var/lib/mailman/cgi-bin/admin %request Accept: image/gif, image/x-xbitmap, image/jpeg, image/pjpeg, image/png, */* Accept-Charset: iso-8859-1,*,utf-8 Accept-Encoding: gzip Accept-Language: en Connection: Keep-Alive Content-length: 40 Content-type: application/x-www-form-urlencoded Cookie: mud-dev:admin="(lp1\012S'216.148.243.2'\012p2\012aI952846157\012aI952856957\012aI-96722254\012a." Host: www.kanga.nu Referer: http://www.kanga.nu/lists/admin/library/ User-Agent: Mozilla/4.7 [en] (X11; I; Linux 2.2.13 i686; Nav) adminpw=xxxxx&request_login=Let+me+in... %response %stderr =======40 --<cut>-- According to that Apache is sending all the right data, it just ain't getting there. I've also upgraded libc6 recently (Debian system FWIW with packages libc6_2.1.3-[567]_i386.deb (notice regex range)). I downgraded libc6 to the package versions (5 and 6) that were installed when it did work to no effect. I also rebuilt the Mailman C based CGI programs under the installed version of libc6, and copied them over (in case something odd was going on) to no observed effect. Note: Plopping the cgi.py in as a standalone cgi tests out as working using GET methods (ie encode vars on the URL). I haven't built a framework yet to test POST for that module. I guess the next step is to either build that framework, or test-drive the C CGIs to make sure stdin really has the right bumph sitting there waiting. -- J C Lawrence Home: claw(a)kanga.nu ----------(*) Other: coder(a)kanga.nu --=| A man is as sane as he is dangerous to his environment |=--

2 1

Re: [Mailman-Developers] Posting to 2 lists breaks newsgating for the second one
by Harald Meland March 11, 2000

March 11, 2000

[Jim Hebert] > I filed a jitterbug report about this a couple weeks back... > > Applix is fairly motivated about using mailman for their internal lists > but needs the ability to crosspost on the mail side and have it show up in > all the right newsgroups too. (If this doesn't make sense, visit the > jitterbug report below.) I suspect this has something to do with the fact that two separate Usenet messages can't have the same Message-Id. If you could have a look in your logs/post Mailman log, you might actually find some foundation for my suspicion (due to the except: clause around line 125 of Mailman/Handlers/ToUsenet.py). If this indeed is the problem at hand, I'm not sure how best to fix it. The optimal solution would be if mail "crossposts" actually turned into proper Usenet crossposts, and not multiple (nearly) identical posts to different newsgroups. However, I don't know enough about NNTP to say if implementing something like that is at all possible. A quick band-aid solution might be to change the Message-Id a wee bit (in a way that makes the change differ from list to list) before passing the message to the NNTP server. > So, I got them to put some money behind the bug report, if someone will > fix it: > > http://www.python.org/mailman-bugs/incoming?id=202;user=guest > http://www.cosource.com/cgi-bin/cos.pl/wish/info/295 > > (Sorry if this is better posted to the -users list, I just joined > both lists and don't have a clear sense of which one would have been > better...) I haven't had time to read -users for a looooong time, but I think -developers is the proper place to post specific feature requests like this. > All in all, we're loving mailman! =) Great :) -- Harald

1 0

Re: [Mailman-Developers] Mailman password and administrative reminders
by Harald Meland March 11, 2000

March 11, 2000

[David Birnbaum] > Here's the headers for the reminder: > > Return-Path: <mailman-owner(a)chelsea.net> > Received: from killian.chelsea.net (localhost [127.0.0.1]) > by killian.chelsea.net (8.9.3/8.9.3) with ESMTP id FAA15246 > for <donad(a)gaigo.sundai.ac.jp>; Wed, 1 Mar 2000 05:52:23 -0500 (EST) > Date: Wed, 1 Mar 2000 05:52:23 -0500 (EST) > Message-Id: <200003011052.FAA15246(a)killian.chelsea.net> > From: mailman-owner(a)chelsea.net > Subject: thebody.com mailing list memberships reminder > To: ###A USER### > X-No-Archive: yes > Precedence: bulk > X-Mailman-Version: 1.1 > Precedence: bulk > List-Id: The Test List <test.chelsea.net> > > It should be from (I think), thebody-admin(a)thebody.com, with an Errors-To: > header. > > Not sure why List-Id: is The Test List...that's odd. That's because the script, while iterating through all lists, looks for the first "public" list. When found, that list is remembered, and that list's (fasttrack) delivery pipeline is used for sending out all the messages. As the "test" list is usually created pretty early on in any particular Mailman installation, it is not entirely uncommon that this list's List-Id: gets stuck on all password reminders... Normally it is nice that Mailman indicates which list's fasttrack pipeline you received a particular message through, but I can see that it might be confusing in this situation. However, a proper remedy is complicated by the lack of a list-independent Mailman mail address -- and thus, as Barry mentioned, also by the lack of a user database. -- Harald

2 1