I installed the most recent git pull of rslight.

I configured it with the exact same configuration that was working on
another install via copy pasta.

I fired the cron for the first time. This was output after it
appeared to cycle through checking messages but never pulled any
messages from remote:

- Trace paths truncated -

PHP Fatal error: Uncaught TypeError: array_diff(): Argument #1
($array) must be of type array, bool given in
..../config/scripts/cron.php:154 Stack trace: #0
..../config/scripts/cron.php(154): array_diff() #1
..../config/scripts/cron.php(138): expire_files() #2 {main}
thrown in .../config/scripts/cron.php on line 154

I fired cron about 20 more times with the same result every time. I
switched between php8.2 and php8.1 for the cron job but got the same
results either way.

I reinstalled from scratch and tried again with same results.

Nothing is being written to spool (tradspool setting). Also the nntp
and nntp-ssl servers do not show up in the linux processes at all.

On Thu, 7 Dec 2023 07:21:08 -0600, Syber Shock wrote:

> I installed the most recent git pull of rslight.
>
> I configured it with the exact same configuration that was working on
> another install via copy pasta.
>
> I fired the cron for the first time. This was output after it appeared
> to cycle through checking messages but never pulled any messages from
> remote:
>
> - Trace paths truncated -
>
> PHP Fatal error: Uncaught TypeError: array_diff(): Argument #1 ($array)
> must be of type array, bool given in .../config/scripts/cron.php:154
> Stack trace: #0 .../config/scripts/cron.php(154): array_diff() #1
> .../config/scripts/cron.php(138): expire_files() #2 {main}
> thrown in .../config/scripts/cron.php on line 154

Hmmm, interesting. This is in expire function related to nocem. I have
nocem configured on my test site, but maybe if it's not used, it breaks.

Could you first try to just create the dirs:
$nocem_processed = $spooldir . "/nocem/processed/";
$nocem_failed = $spooldir . "/nocem/failed/";

or, let me know if they already exist? Then need to be readable by web
user.

If that fixes it, I'll modify the function to try to handle that.

On Thu, 7 Dec 2023 13:38:15 +0000
Retro Guy <retroguy@novabbs.org> wrote:

> On Thu, 7 Dec 2023 07:21:08 -0600, Syber Shock wrote:
>
> > I installed the most recent git pull of rslight.
> >
> > I configured it with the exact same configuration that was working
> > on another install via copy pasta.
> >
> > I fired the cron for the first time. This was output after it
> > appeared to cycle through checking messages but never pulled any
> > messages from remote:
> >
> > - Trace paths truncated -
> >
> > PHP Fatal error: Uncaught TypeError: array_diff(): Argument #1
> > ($array) must be of type array, bool given in
> > .../config/scripts/cron.php:154 Stack trace: #0
> > .../config/scripts/cron.php(154): array_diff() #1
> > .../config/scripts/cron.php(138): expire_files() #2 {main} thrown
> > in .../config/scripts/cron.php on line 154
>
> Hmmm, interesting. This is in expire function related to nocem. I have
> nocem configured on my test site, but maybe if it's not used, it
> breaks.
>
> Could you first try to just create the dirs:
> $nocem_processed = $spooldir . "/nocem/processed/";
> $nocem_failed = $spooldir . "/nocem/failed/";
>
> or, let me know if they already exist? Then need to be readable by
> web user.
>
> If that fixes it, I'll modify the function to try to handle that.

Creating the directories manually solved the type error. Now the cron
completes without throwing any error codes. But this creates a new
problem.

I ran about another 25 times and the only group that gets written to
spool/articles is rocksolid.spam. No folders or files are created for
any of the other groups.

Only 4 articles show up in rocksolid.spam even after many runs of the
cron job.

Also, the web interface updated one time and populated the list of all
the rocksolid groups. Then I removed all but three of the groups and
the web interface never updated even after many runs.

I checked the permissions several times on the web, config, and spool
files. Everything seems fine there.

Syber Shock wrote:

> On Thu, 7 Dec 2023 13:38:15 +0000
> Retro Guy <retroguy@novabbs.org> wrote:

>> On Thu, 7 Dec 2023 07:21:08 -0600, Syber Shock wrote:
>>
>> > I installed the most recent git pull of rslight.
>> >
>> > I configured it with the exact same configuration that was working
>> > on another install via copy pasta.
>> >
>> > I fired the cron for the first time. This was output after it
>> > appeared to cycle through checking messages but never pulled any
>> > messages from remote:
>> >
>> > - Trace paths truncated -
>> >
>> > PHP Fatal error: Uncaught TypeError: array_diff(): Argument #1
>> > ($array) must be of type array, bool given in
>> > .../config/scripts/cron.php:154 Stack trace: #0
>> > .../config/scripts/cron.php(154): array_diff() #1
>> > .../config/scripts/cron.php(138): expire_files() #2 {main} thrown
>> > in .../config/scripts/cron.php on line 154
>>
>> Hmmm, interesting. This is in expire function related to nocem. I have
>> nocem configured on my test site, but maybe if it's not used, it
>> breaks.
>>
>> Could you first try to just create the dirs:
>> $nocem_processed = $spooldir . "/nocem/processed/";
>> $nocem_failed = $spooldir . "/nocem/failed/";
>>
>> or, let me know if they already exist? Then need to be readable by
>> web user.
>>
>> If that fixes it, I'll modify the function to try to handle that.

> Creating the directories manually solved the type error. Now the cron
> completes without throwing any error codes. But this creates a new
> problem.

> I ran about another 25 times and the only group that gets written to
> spool/articles is rocksolid.spam. No folders or files are created for
> any of the other groups.

> Only 4 articles show up in rocksolid.spam even after many runs of the
> cron job.

> Also, the web interface updated one time and populated the list of all
> the rocksolid groups. Then I removed all but three of the groups and
> the web interface never updated even after many runs.

> I checked the permissions several times on the web, config, and spool
> files. Everything seems fine there.

I'll need to set up a tradspool install to see what's going on.

I'll fix the missing dir problem (nocem), and I will do a tradspool install. I assume everything is set up correctly in menu.conf, <section>:1:1

I have made changes to spoolnews.php recently, so I may have broken something. Once I have a tradspool install, I'll keep it. I'll get on this tomorrow and let you know what I find.

--
Retro Guy

On Thu, 7 Dec 2023 22:07:39 +0000
retro.guy@rocksolidbbs.com (Retro Guy) wrote:

> Syber Shock wrote:
>
> > On Thu, 7 Dec 2023 13:38:15 +0000
> > Retro Guy <retroguy@novabbs.org> wrote:
>
> >> On Thu, 7 Dec 2023 07:21:08 -0600, Syber Shock wrote:
> >>
> >> > I installed the most recent git pull of rslight.
> >> >
> >> > I configured it with the exact same configuration that was
> >> > working on another install via copy pasta.
> >> >
> >> > I fired the cron for the first time. This was output after it
> >> > appeared to cycle through checking messages but never pulled any
> >> > messages from remote:
> >> >
> >> > - Trace paths truncated -
> >> >
> >> > PHP Fatal error: Uncaught TypeError: array_diff(): Argument #1
> >> > ($array) must be of type array, bool given in
> >> > .../config/scripts/cron.php:154 Stack trace: #0
> >> > .../config/scripts/cron.php(154): array_diff() #1
> >> > .../config/scripts/cron.php(138): expire_files() #2 {main} thrown
> >> > in .../config/scripts/cron.php on line 154
> >>
> >> Hmmm, interesting. This is in expire function related to nocem. I
> >> have nocem configured on my test site, but maybe if it's not used,
> >> it breaks.
> >>
> >> Could you first try to just create the dirs:
> >> $nocem_processed = $spooldir . "/nocem/processed/";
> >> $nocem_failed = $spooldir . "/nocem/failed/";
> >>
> >> or, let me know if they already exist? Then need to be readable by
> >> web user.
> >>
> >> If that fixes it, I'll modify the function to try to handle that.
>
> > Creating the directories manually solved the type error. Now the
> > cron completes without throwing any error codes. But this creates a
> > new problem.
>
> > I ran about another 25 times and the only group that gets written to
> > spool/articles is rocksolid.spam. No folders or files are created
> > for any of the other groups.
>
> > Only 4 articles show up in rocksolid.spam even after many runs of
> > the cron job.
>
> > Also, the web interface updated one time and populated the list of
> > all the rocksolid groups. Then I removed all but three of the
> > groups and the web interface never updated even after many runs.
>
> > I checked the permissions several times on the web, config, and
> > spool files. Everything seems fine there.
>
> I'll need to set up a tradspool install to see what's going on.

I also set up with database instead of tradspool and got the same
problems with every new install, to wit:

1. No messages are synced, at all, nada, with the exception of the
first time when 4 spam messages synced.

2. It doesn't even look like the remote server is ever contacted or
connected to, even though the cron output throws no error.
> I'll fix the missing dir problem (nocem), and I will do a tradspool
> install. I assume everything is set up correctly in menu.conf,
> <section>:1:1

Yes, everything is set up right. I didn't touch the default install for
several tests, just entered peer auth and address, and still have the
same issues.

> I have made changes to spoolnews.php recently, so I may have broken
> something. Once I have a tradspool install, I'll keep it. I'll get on
> this tomorrow and let you know what I find.

One more issue is the confusion around the 'remote_ssl' param in
rslight.config.php. In earlier versions the user had to put a '1' here
to enable ssl, then the port in 'remote_port'. The config.php setup
still shows this instruction, which conflicts with how it actually
works now.

Happy bug hunting. I'm plum tuckered out.

Syber Shock wrote:

> On Thu, 7 Dec 2023 22:07:39 +0000
> retro.guy@rocksolidbbs.com (Retro Guy) wrote:

>> Syber Shock wrote:
>>
>> > On Thu, 7 Dec 2023 13:38:15 +0000
>> > Retro Guy <retroguy@novabbs.org> wrote:
>>
>> >> On Thu, 7 Dec 2023 07:21:08 -0600, Syber Shock wrote:
>> >>

<snip>
>>
>> I'll need to set up a tradspool install to see what's going on.

> I also set up with database instead of tradspool and got the same
> problems with every new install, to wit:

> 1. No messages are synced, at all, nada, with the exception of the
> first time when 4 spam messages synced.

> 2. It doesn't even look like the remote server is ever contacted or
> connected to, even though the cron output throws no error.

This is very odd. I downloaded the .zip from gitlab and installed in a vm. The only change I made was to add the remote server, user/password to the config. No other changes.

It's syncing, and I'm posting from it right now.

> Yes, everything is set up right. I didn't touch the default install for
> several tests, just entered peer auth and address, and still have the
> same issues.

I'm going to try the same as above (fresh install from .zip) with tradspool and report the results.

> One more issue is the confusion around the 'remote_ssl' param in
> rslight.config.php. In earlier versions the user had to put a '1' here
> to enable ssl, then the port in 'remote_port'. The config.php setup
> still shows this instruction, which conflicts with how it actually
> works now.

Once we get you up and running, I'll get this other stuff taken care of.

> Happy bug hunting. I'm plum tuckered out.

I always appreciate the bug reports and feedback. Get some rest now! :)

--
Unconfigured rslight site

<snip> continuation of thread as bug report data </snip>

Retro Guy: will you check your NNTP access logs for logs containing user
'sybershock' at i2pn2.org and novabbs.org?

Some months ago I ran a few older versions of rslight on this server. I
had version 7_2 running for a long time. I had version 9_1
running at one point but with some problems that I forget exactly.

I tried again with the most recent rslight version set for tradspool.
Newsportal and NNTP server never come online. Data below.

OS info:

OS: Debian GNU/Linux 12 (bookworm) x86_64
Kernel: 6.1.0-13-amd64

Server info:

Package: nginx
Version: 1.25.3-1~bookworm

Package: apache2
Version: 2.4.58-1+0~20231025.34+debian12~1.gbp0fd694

nginx is proxy with php-fpm enabled.

Steps taken:

1. Download .zip from gitlab.
2. Unpack .zip.
3. Run installer and mkdir spool/nocem/failed & spool/nocem/processed.
4. Configured using /common/setup.php for test account at i2pn2.org.
5. Manually connected to i2pn2.org with openssl.
6. Verified authinfo and got successful authentication at i2pn2.org.
==> 281 Authentication succeeded
7. Fired the cron job manually by copypasting the output from installer.
8. Looped the cron job many times with short intervals.
9. Each run of the cron job gave exactly this output:

Updated user count
Sending articles
Send Done
Refreshed spoolnews
Expired articles
Sending articles
Send Done
Refreshed spoolnews
Expired articles
RSS Feeds updated
Log files rotated
Keys rotated
Removed old files

There are no exceptions to this output. It is exactly the same output on
every cron run during testing.

10. PHP portal pages give this error on all tries:
==> Error: The connection to the Message Server failed.
11. The cron job never activates newsportal.php.
12. The local NNTP server never runs at all.
13. Checked firewall and NNTP ports are whitelisted.
14. 100% certain that the NNTP server is never activated by cron job.
15. No nntp.lock or nntp-ssl.lock ever appears in spool/lock folder.
16. Contents of spool/lock folder:
==> ls >>> rocksolid-spoolnews.lock spoolnews-spoolnews.lock
17. Contents of spool folder:

articles-overview.db3
keys.dat
lock
log
nocem
rocksolid
rocksolid.nodes.announce-data.db3
rocksolid.nodes-data.db3
rocksolid.nodes.help-data.db3
rocksolid.programming-data.db3
rocksolid.shared.encryption-data.db3
rocksolid.shared.entertainment-data.db3
rocksolid.shared.freenet-data.db3
rocksolid.shared.general-data.db3
rocksolid.shared.hacking-data.db3
rocksolid.shared.helpdesk-data.db3
rocksolid.shared.i2p-data.db3
rocksolid.shared.linux-data.db3
rocksolid.shared.news-data.db3
rocksolid.shared.offtopic-data.db3
rocksolid.shared.rocksolid-data.db3
rocksolid.shared.security-data.db3
rocksolid.shared.test-data.db3
rocksolid.shared.tor-data.db3
rocksolid.social-data.db3
rocksolid.spam-data.db3
rocksolid-thread-timer
sessions.dat
spoolnews
spoolnews-thread-timer
ssl
upload

18. <?php phpinfo(); ?> [truncated the irrelevant bits]

System Linux sybershock.com 6.1.0-13-amd64 #1 SMP PREEMPT_DYNAMIC
Build Date Nov 24 2023 13:10:42
Build System Linux
Server API FPM/FastCGI
Virtual Directory Support disabled
Configuration File (php.ini) Path /etc/php/8.2/fpm
Loaded Configuration File /etc/php/8.2/fpm/php.ini
Scan this dir for additional .ini files /etc/php/8.2/fpm/conf.d
Debian 6.1.55-1 (2023-09-29) x86_64

Additional .ini files parsed 10-mysqlnd.ini, 10-opcache.ini,
10-pdo.ini, 15-xml.ini, 20-bcmath.ini, 20-bz2.ini, 20-calendar.ini,
20-ctype.ini, 20-curl.ini, 20-dom.ini, 20-exif.ini, 20-ffi.ini,
20-fileinfo.ini, 20-ftp.ini, 20-gd.ini, 20-gettext.ini, 20-gmp.ini,
20-gnupg.ini, 20-iconv.ini, 20-imagick.ini, 20-imap.ini, 20-intl.ini,
20-ldap.ini, 20-mbstring.ini, 20-mysqli.ini, 20-pdo_mysql.ini,
20-pdo_sqlite.ini, 20-phar.ini, 20-posix.ini, 20-pspell.ini,
20-readline.ini, 20-shmop.ini, 20-simplexml.ini, 20-soap.ini,
20-sockets.ini, 20-sqlite3.ini, 20-sysvmsg.ini, 20-sysvsem.ini,
20-sysvshm.ini, 20-tokenizer.ini, 20-xmlreader.ini, 20-xmlwriter.ini,
20-xsl.ini, 20-zip.ini

GD Support enabled
GPGme Version 1.18.0
hash support enabled
Multibyte Support (mbstring) enabled
OpenSSL support enabled
PCRE (Perl Compatible Regular Expressions) Support enabled
PDO drivers mysql, sqlite
Session Support enabled
Sockets Support enabled
SQLite Library 3.40.1

19. PHP info shows rslight config/ & spool/ in open_basedir config.
20. No entries in /var/log/nginx/domains/ error log for domain.
21. No entries for domain in any error logs in /var/log/nginx.
22. Files created in spool/log# ls

access.log access.log.1 cron.log cron.log.1 expire.log
expire.log.1 nntp.log.1 rotate spoolnews.log spoolnews.log.1

23. Cat nntp.log.1
==> Dec 08 21:03:40 Creating self-signed certificate

24. spool/log# cat access.log

Dec 08 21:14:15 forum.sybershock.com/rocksolid/index.php
Dec 08 21:20:24 forum.sybershock.com/rocksolid/index.php
Dec 08 21:50:09 forum.sybershock.com/rocksolid/index.php
Dec 08 21:59:41 forum.sybershock.com/rocksolid/index.php
Dec 08 21:59:43 forum.sybershock.com/rocksolid/?
Dec 08 21:59:45 forum.sybershock.com/rocksolid/overboard.php?
Dec 08 21:59:50 forum.sybershock.com/rocksolid/index.php
Dec 08 21:59:51 forum.sybershock.com/rocksolid/search.php?
Dec 08 21:59:57 forum.sybershock.com/rocksolid/search.php
Dec 08 22:00:03 forum.sybershock.com/spoolnews/mail.php
Dec 08 22:00:04 forum.sybershock.com/spoolnews/files.php
Dec 08 22:00:10 forum.sybershock.com/spoolnews/user.php
Dec 08 22:00:12 forum.sybershock.com/rocksolid/index.php
Dec 08 22:00:16 forum.sybershock.com/rocksolid/index.php

25. log# cat access.log.1
==> Dec 08 20:58:47 forum.sybershock.com/rocksolid/index.php

26. spool/log# cat cron.log.1
==> Dec 08 21:03:40 spoolnews cron 78224 started...

27. expire.log has hundreds of VACUUM entries.

I tried alternate configurations of rslight.config.php:

-- 1 --

'remote_server' => 'news.i2pn2.org',
'remote_port' => '119',
'remote_ssl' => '',

-- 2 --

'remote_server' => 'news.i2pn2.org',
'remote_port' => '563',
'remote_ssl' => '1',

-- 3 --

'remote_server' => 'news.i2pn2.org',
'remote_port' => '563',
'remote_ssl' => '563',

-- 4 --

'remote_server' => 'news.i2pn2.org',
'remote_port' => '',
'remote_ssl' => '563',

-- 5 --

'remote_server' => 'news.i2pn2.org',
'remote_port' => '119',
'remote_ssl' => '563',

None of these configurations work or give any error output.

28. spool/log# cat spoolnews.log

Dec 08 21:07:20 spoolnews remote_server: news.i2pn2.org
Dec 08 21:08:24 spoolnews remote_server: news.i2pn2.org
Dec 08 21:09:41 spoolnews remote_server: news.i2pn2.org
[truncated]
Dec 08 22:15:28 spoolnews remote_server: news.i2pn2.org
Dec 08 22:15:42 spoolnews remote_server: news.i2pn2.org
Dec 08 22:15:56 spoolnews remote_server: news.i2pn2.org
Dec 08 22:16:09 spoolnews remote_server: news.i2pn2.org

I also get the same output for all other servers tried.

29. cat cron.log

Dec 08 21:07:14 spoolnews cron 78746 started...
Dec 08 21:07:35 spoolnews cron 78746 completed...
[truncated]
Dec 08 22:19:04 spoolnews cron 90219 started...
Dec 08 22:19:07 spoolnews cron 90219 completed...

30. spool/lock# tai1 * 2 *
==> rocksolid-spoolnews.lock <==
90265
==> spoolnews-spoolnews.lock <==
90259

31. I ran the cron job manually several times with several different
versions of php: php7.4, php8.0, php8.1, php8.2, php8.3.

After dozens of runs of the cron job the php error log is
still an empty file. The cron shows no errors in output. No messages
are ever synced and nothing is added to the spool. The local NNTP
server never runs and never shows up in the Linux process manager.

All I can tell is that the local newsportal NNTP server never runs at
all and no errors of any kind appear anywhere I would expect errors.

After all of this the only error ever generated was a page error:

[Fri Dec 08 17:00:08.666811 2023] [proxy_fcgi:error] [pid 9078:tid
139943439423168] [client 174.247.224.6:0] AH01071: Got error 'PHP
message: PHP Warning: Undefined variable $title in .../common/head.inc
on line 4; PHP message: PHP Warning: Undefined variable $config_dir in
..../common/head.inc on line 8; PHP message: PHP Warning:
file_exists(): open_basedir restriction in effect.
File(/googleanalytics.conf) is not within the allowed path(s):
[truncated] in .../common/head.inc on line 8', referer:
https://forum.sybershock.com/common/register.php

<snip> continuation of thread as bug report data </snip>

So far in this adventure this is the only error output to the site
error log:

[Fri Dec 08 21:00:30.241715 2023] [proxy_fcgi:error] [pid 25642:tid
139703332300480] [client 174.247.224.6:0] AH01071: Got error 'PHP
message: PHP Fatal error: Uncaught Error: Call to undefined function
fdiv() in .../rocksolid/newsportal.php:1796\nStack trace:\n#0
..../rocksolid/head.inc(10): throttle_hits()\n#1
..../rocksolid/index.php(22): include('/home/nightbulb...')\n#2 {main}\n
thrown in .../rocksolid/newsportal.php on line 1796'

And I believe this error may be from the front end rather than the
cron job.

I ran another re-install this time installing all the spool and config
in /opt/rslight and it still gives the same errorless failures as
before with php8.2 and php7.4 on backend and cron. Newsportal never
runs no matter how many times I run the cron job. And the NNTP server
never starts up and no articles are written to spool. The 'articles'
directory doesn't get created. The size of the data files stays
constant as if no data is being pulled from remote. I combed the
php, apache, and domain logs and syslog and could not find anything
relevant.

I even disabled SSL and used the unencrypted connection on port 119 for
remote host. I changed the password right quick after that run.

I installed rslight on another debian machine and I was able to
syncronize messages via Tor with that fresh install using the same
configuration in rslight.inc.php.

I am at my wit's end. I cannot find anything to indicate what the
problem is. Got any ideas?

I wonder ...

$ cgi-fcgi -bind -connect 127.0.0.1:1119
Could not connect to 127.0.0.1:1119

But netcat can grab the port:

$ nc -l 1119

Is it possible that somewhere in the guts of php-fpm that listening on
ports is blocked or locked such that the newsportal module would fail to
register a port with silent failure? Where might I look?

On another fresh install I decided to run the installer as user instead
of root. Some interesting errors occurred. First, the installer output
sed errors:

Moving files into place...done

Setting permissions...done

Applying configuration...sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: couldn't open temporary file /opt/rslight/config/sedF55cUn:
Permission denied sed: can't read : No such file or directory
sed: couldn't open temporary file /opt/rslight/config/sedkxr6sf:
Permission denied sed: can't read : No such file or directory
sed: couldn't open temporary file /opt/rslight/config/sed0TqTjt:
Permission denied sed: can't read : No such file or directory
sed: couldn't open temporary file /opt/rslight/config/seddxn7mp:
Permission denied sed: can't read : No such file or directory
sed: couldn't open temporary file /opt/rslight/config/sedhGucpD:
Permission denied sed: can't read : No such file or directory
sed: couldn't open temporary file /opt/rslight/config/sedHCZJXm:
Permission denied sed: can't read : No such file or directory
sed: couldn't open temporary file /opt/rslight/config/sedYTsubW:
Permission denied done

Secondly, when I entered the config directory I found this:

$ cat admin.inc.php
<?php
$admin = array();
$admin['password'] = '<admin_password>';
$admin['key'] = '<admin_key>';
?>

Thirdly, I had to use the actual text, '<admin_password>', to log in to
/common/setup.php.

Then I re-ran the debian-install.sh script as root and got less explicit
errors with the sed commands:

Applying configuration...sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
sed: can't read : No such file or directory
done

This time it gave a proper, randomly generated key in admin.inc.php.

Then I configured the remote auth data in rslight.inc.php and I ran the
cron:

Updated user count
Sending articles
Send Done
Refreshed spoolnews
Expired articles
Sending articles
Send Done
Refreshed spoolnews
Expired articles
RSS Feeds updated
Rotated: /opt/rslight/spool//log/access.log
Rotated: /opt/rslight/spool//log/cron.log
Rotated: /opt/rslight/spool//log/expire.log
Rotated: /opt/rslight/spool//log/nntp.log
Rotated: /opt/rslight/spool//log/spoolnews.log
Log files rotated
Keys rotated
Removed old files

Double slashes show up here but I did not use trailing slashes for the
path datas on the installer wizard.

And still the NNTP server does not start or show up in linux process
manager and no remote data is fetched into the /spool directory. There
are no errors to be found anywhere. I even read over 3000 lines of
strace output yet I did not find anything relevant in it.

I wanted to check what is happening between the local rslight install
and the remote server during the cron job. So I used a proxy to dump the
session to file while the cron was running. Here is the complete
content of the session.

200 rocksolid2 InterNetNews NNRP server INN 2.7.1 ready (no posting)
MODE reader
200 rocksolid2 InterNetNews NNRP server INN 2.7.1 ready (no posting)
AUTHINFO USER [redacted]
381 Enter password
AUTHINFO PASS [redacted]
281 Authentication succeeded
QUIT
200 rocksolid2 InterNetNews NNRP server INN 2.7.1 ready (no posting)
MODE reader
200 rocksolid2 InterNetNews NNRP server INN 2.7.1 ready (no posting)
AUTHINFO USER [redacted]
381 Enter password
AUTHINFO PASS [redacted]
281 Authentication succeeded
200 rocksolid2 InterNetNews NNRP server INN 2.7.1 ready (no posting)
MODE reader
200 rocksolid2 InterNetNews NNRP server INN 2.7.1 ready (no posting)
AUTHINFO USER [redacted]
381 Enter password
AUTHINFO PASS [redacted]
281 Authentication succeeded
QUIT

It seems like it authenticates then starts over a few times without
sending any other commands.

Then I ran a manual session and responded by typing responses to the
local rslight cron job, and I got the same loop.

On Fri, 8 Dec 2023 16:51:40 -0600, Syber Shock wrote:

> <snip> continuation of thread as bug report data </snip>
>
> Retro Guy: will you check your NNTP access logs for logs containing user
> 'sybershock' at i2pn2.org and novabbs.org?

I'm seeing logins, but no activity. (Just log in and leave).

> I tried again with the most recent rslight version set for tradspool.
> Newsportal and NNTP server never come online. Data below.
>
> OS info:
> OS: Debian GNU/Linux 12 (bookworm) x86_64 Kernel: 6.1.0-13-amd64
> Server info:
> Package: nginx Version: 1.25.3-1~bookworm
> Package: apache2 Version: 2.4.58-1+0~20231025.34+debian12~1.gbp0fd694
> nginx is proxy with php-fpm enabled.

Thanks for this info. My testing system:
OS: Debian 12 (Bookworm) x86_64 Kernel: 6.1.37-1
Nginx: 1.22.1-9
PHP, php-fpm: 8.2.7

Install dirs: DEFAULT (/etc/rslight, /var/spool/rslight, /var/www/html)
All suggested packages from INSTALL file installed.

> Steps taken:
>
> 1. Download .zip from gitlab.
> 2. Unpack .zip.
> 3. Run installer and mkdir spool/nocem/failed & spool/nocem/processed.

I find that I do not need to create these dirs manually. They are only
created if 'nocem' is configured to be used, but I get NO errors without
these dirs existing.

> 4. Configured using /common/setup.php for test account at i2pn2.org.
> 5. Manually connected to i2pn2.org with openssl.
> 6. Verified authinfo and got successful authentication at i2pn2.org.
> ==> 281 Authentication succeeded

YAY!

> 7. Fired the cron job manually by copypasting the output from installer.

You may try (this is how I do it) to run it manually by:
cd <$webdir>/spoolnews
php <$config_dir>/scripts/cron.php

Run the above as 'root' user (cron.php must be run as root)

> 8. Looped the cron job many times with short intervals.
> 9. Each run of the cron job gave exactly this output:
>
> Updated user count Sending articles Send Done Refreshed spoolnews
> Expired articles Sending articles Send Done Refreshed spoolnews Expired
> articles RSS Feeds updated Log files rotated Keys rotated Removed old
> files
>
> There are no exceptions to this output. It is exactly the same output on
> every cron run during testing.

Interesting. It is making it through the entire script.

> 10. PHP portal pages give this error on all tries:
> ==> Error: The connection to the Message Server failed.
> 11. The cron job never activates newsportal.php.
> 12. The local NNTP server never runs at all.
> 13. Checked firewall and NNTP ports are whitelisted.
> 14. 100% certain that the NNTP server is never activated by cron job.
> 15. No nntp.lock or nntp-ssl.lock ever appears in spool/lock folder.

I know you know this, but just putting it out there in case:
'local_server' => '[::]', // where to bind
'local_port' => '1119', // port to bind
'local_ssl_port' => '1563', // ssl port to bind if using ssl
'enable_all_networks' => '', // just bind to all available
'server_auth_user' => 'localuser', // local nntp server username for
newsportal
'server_auth_pass' => 'blahblahblah', // local nntp server password for
newsportal

> 23. Cat nntp.log.1
> ==> Dec 08 21:03:40 Creating self-signed certificate

This is actually run from cron.php

>
> 27. expire.log has hundreds of VACUUM entries.

expire should actually create a file in $spooldir to keep track of when it
last ran.
You should see files like 'rocksolid-expire-timer' in $spooldir, and when
you try to run expire.php again, it should immediately quit if the file is
not at least 24 hours old.

Are these files being created?
>
> 28. spool/log# cat spoolnews.log
>
> Dec 08 21:07:20 spoolnews remote_server: news.i2pn2.org Dec 08 21:08:24
> spoolnews remote_server: news.i2pn2.org Dec 08 21:09:41 spoolnews
> remote_server: news.i2pn2.org [truncated]
> Dec 08 22:15:28 spoolnews remote_server: news.i2pn2.org Dec 08 22:15:42
> spoolnews remote_server: news.i2pn2.org Dec 08 22:15:56 spoolnews
> remote_server: news.i2pn2.org Dec 08 22:16:09 spoolnews remote_server:
> news.i2pn2.org
>
> I also get the same output for all other servers tried.

Are the config files in $config_dir (mainly groups.txt in sections)
readable by www-data (or whatever www user)?
Is menu.conf default and readable?

> All I can tell is that the local newsportal NNTP server never runs at
> all and no errors of any kind appear anywhere I would expect errors.

Try (as root):
cd $webdir/spoolnews
php $config_dir/scripts/nntp.php

and see what errors you get (if any).

> [Fri Dec 08 21:00:30.241715 2023] [proxy_fcgi:error] [pid 25642:tid
> 139703332300480] [client 174.247.224.6:0] AH01071: Got error 'PHP
> message: PHP Fatal error: Uncaught Error: Call to undefined function
> fdiv() in .../rocksolid/newsportal.php:1796\nStack trace:\n#0
> .../rocksolid/head.inc(10): throttle_hits()\n#1
> .../rocksolid/index.php(22): include('/home/nightbulb...')\n#2 {main}\n
> thrown in .../rocksolid/newsportal.php on line 1796'
>
> And I believe this error may be from the front end rather than the cron
> job.

fdiv() is in php8. You'll see the error in php7, but should not in php8.
This function is used to avoid a possible division by zero error.

Yes, it is from the newsportal part.

>
> I installed rslight on another debian machine and I was able to
> syncronize messages via Tor with that fresh install using the same
> configuration in rslight.inc.php.

So it worked on another machine? What is different between the systems
(I'm sure you thought about this already :)

> I wonder ...
>
> $ cgi-fcgi -bind -connect 127.0.0.1:1119 Could not connect to
> 127.0.0.1:1119
>
> But netcat can grab the port:
>
> $ nc -l 1119

Do you see nntp.php running with 'ps aux | grep nntp' ?

> Is it possible that somewhere in the guts of php-fpm that listening on
> ports is blocked or locked such that the newsportal module would fail to
> register a port with silent failure? Where might I look?

It's possible, but I am not finding that issue myself with php8. If I
figure out how to check that in php settings, I'll let you know.

> On another fresh install I decided to run the installer as user instead
> of root. Some interesting errors occurred. First, the installer output
> sed errors:
<snip>
> /opt/rslight/config/sed0TqTjt: Permission denied sed: can't read : No
> such file or directory sed: couldn't open temporary file
<snip>
>
> Secondly, when I entered the config directory I found this:
>
> $ cat admin.inc.php <?php
> $admin = array();
> $admin['password'] = '<admin_password>';
> $admin['key'] = '<admin_key>';
> ?>

That is what's in that file before sed operates on it. It's how sed knows
where to put the username and password.

>
> Then I re-ran the debian-install.sh script as root and got less explicit
> errors with the sed commands:
>
> Applying configuration...sed: can't read : No such file or directory
> sed: can't read : No such file or directory sed: can't read : No such
> file or directory sed: can't read : No such file or directory sed: can't
> read : No such file or directory sed: can't read : No such file or
> directory sed: can't read : No such file or directory sed: can't read :
> No such file or directory sed: can't read : No such file or directory
> sed: can't read : No such file or directory done

The above is a known issue. Error (can't read) but it works anyway. Not
sure about that one (it works in BSD)

> Updated user count Sending articles Send Done Refreshed spoolnews
> Expired articles Sending articles Send Done Refreshed spoolnews Expired
> articles RSS Feeds updated Rotated: /opt/rslight/spool//log/access.log
> Rotated: /opt/rslight/spool//log/cron.log Rotated:
> /opt/rslight/spool//log/expire.log Rotated:
> /opt/rslight/spool//log/nntp.log Rotated:
> /opt/rslight/spool//log/spoolnews.log Log files rotated Keys rotated
> Removed old files
>
> Double slashes show up here but I did not use trailing slashes for the
> path datas on the installer wizard.

Double slashes should not be a problem. the OS handles these like single.
They are there in case an admin forgets or removes a trailing slash in
config files, so that it will work anyway.

> And still the NNTP server does not start or show up in linux process
> manager and no remote data is fetched into the /spool directory. There
> are no errors to be found anywhere. I even read over 3000 lines of
> strace output yet I did not find anything relevant in it.

I finally got some error output from a fresh install of rslight.

I cloned the most recent gitlab repo of rslight.

I double-checked to ensure the php-fpm server backend to uses php8.2.
Still nginx + apache2.

I ran the install using the default path targets for spool and config.

The web root is under a domain name path.

I configured rslight.inc.php with credentials for news.i2pn2.org. For
this test I set it to use database instead of tradspool.

The subsequent runs of the cron job gave the same error output.

The newsportal NNTP server does not activate or run, as before.

The .db3 files are still created in /spool but nothing is changing with
them and no data incoming.

The web portal is inaccessible and gives http error 500 for index.php
and rocksolid directory, but /common/setup.php is accessible.

<snip> another test install with some error copypasta </snip>

I used apt and hestia to remove all versions and vestiges of php prior
to php8.2. I also cleaned apt packages and rebooted after unistalling
cruft. Only php8.2 is available to any user on the machine now.

nntp.php throws an error:

---------------------------------------------------------------------
public_html/spoolnews# php /etc/rslight/scripts/nntp.php
PHP Fatal error: Uncaught Error: Call to undefined function
pcntl_fork() in /etc/rslight/scripts/rslight-lib.php:236 Stack trace:
#0 /etc/rslight/scripts/nntp.php(19): become_daemon()
#1 {main}
thrown in /etc/rslight/scripts/rslight-lib.php on line 236
---------------------------------------------------------------------

Default install still giving same problem. No errors of any kind to
find but no data moves, and no NNTP server startup. Now we know the
NNTP startup problem is related to become_daemon().

I suspect if we figure out how to fix this all the other problems will
be easier to resolve.

And yes, with every test install I was looping ps aux commands and
grepping for nntp and never once did nntp or nntp-ssl ever appear in
the process manager output. I suppose the daemon would need to succeed
before the script would be listed there ...

I tried another install, this time with /config and /spool in
/opt/rocksolid/ instead of the regular locations. I went to the public
spoolnews directory and ran nntp.php and got the same error:

---------------------------------------------------------------------
public_html/spoolnews# php /opt/rocksolid/config/scripts/nntp.php
PHP Fatal error: Uncaught Error: Call to undefined function
pcntl_fork() in /opt/rocksolid/config/scripts/rslight-lib.php:236 Stack
trace: #0 /opt/rocksolid/config/scripts/nntp.php(19): become_daemon()
#1 {main}
thrown in /opt/rocksolid/config/scripts/rslight-lib.php on line 236
---------------------------------------------------------------------

I had to manually create the nocem directories each install to keep the
cron job from crashing with this:

---------------------------------------------------------------------
Rotated: /var/spool/rslight//log/cron.log
Rotated: /var/spool/rslight//log/expire.log
Rotated: /var/spool/rslight//log/nntp.log
Rotated: /var/spool/rslight//log/spoolnews.log
Log files rotated
Keys rotated
PHP Fatal error: Uncaught TypeError: array_diff(): Argument #1
($array) must be of type array, bool given in
/etc/rslight/scripts/cron.php:154 Stack trace: #0
/etc/rslight/scripts/cron.php(154): array_diff() #1
/etc/rslight/scripts/cron.php(138): expire_files() #2 {main}
thrown in /etc/rslight/scripts/cron.php on line 154
---------------------------------------------------------------------

Let me know if you have any ideas on how I can fix the NNTP daemon.

On Sat, 9 Dec 2023 07:34:26 -0600
Syber Shock <3883@sugar.bug> wrote:

> Let me know if you have any ideas on how I can fix the NNTP daemon.

I might have answered my own query. pcntl_fork is disabled:

php -i | grep pcntl
disable_functions => pcntl_alarm,pcntl_fork ...

I'm going to fiddle with this and test again.

On Sat, 9 Dec 2023 07:34:26 -0600, Syber Shock wrote:

> <snip> another test install with some error copypasta </snip>
>
> I used apt and hestia to remove all versions and vestiges of php prior
> to php8.2. I also cleaned apt packages and rebooted after unistalling
> cruft. Only php8.2 is available to any user on the machine now.
>
> nntp.php throws an error:
>
> ---------------------------------------------------------------------
> public_html/spoolnews# php /etc/rslight/scripts/nntp.php
> PHP Fatal error: Uncaught Error: Call to undefined function
> pcntl_fork() in /etc/rslight/scripts/rslight-lib.php:236 Stack trace: #0
> /etc/rslight/scripts/nntp.php(19): become_daemon()
> #1 {main}
> thrown in /etc/rslight/scripts/rslight-lib.php on line 236
> ---------------------------------------------------------------------
>
Any chance pcntl is disabled?

/etc/php/8.2/fpm/php.ini
; This directive allows you to disable certain functions.
; It receives a comma-delimited list of function names.
; https://php.net/disable-functions
disable_functions = <IS 'pcntl_fork' HERE?>

On Sat, 9 Dec 2023 13:51:17 +0000
Retro Guy <retroguy@novabbs.org> wrote:

<snip>

> Any chance pcntl is disabled?

100% chance ;)
<snip>

I need to make a workaround. Brainstorm time.

Have you ever used the 'system daemon' package from PEAR?

"System_Daemon is a PHP class that allows developers to create their
own daemon applications on Linux systems. The class is focussed
entirely on creating & spawning standalone daemons ..."

https://pear.php.net/package/System_Daemon

I am thinking of adding some alternative daemon options because managing
templates for a moving php.ini target (hestia) is a pain. It broke two
other installs of rslight when the system went through several
automatic update cycles. Updates tend to break nice templates, which
makes me fume, especially as I forgot why I made the templates just so.

I'm considering three alternative options for the NNTP daemon:

1. PEAR System Daemon;
2. Init & Bash nohup wrapper daemon;
3. SystemD service.

In fact I am thinking of all three being bundled options. The bash
installer could set up a init or systemd unit that automatically
respawns on crash or failure. It could also check if the PEAR package
is installed and the function calling the NNTP server could case bool
to the PEAR package. Finally it could check the pcntl_fork function but
any failure of it is a FATAL in php. This is another reason I do not
prefer it.

What other hacks haven't I thought of? There have got to be some clever
workarounds that don't require but a few function tweaks ... unless we
have a whole bunch more pcntl functions being used that I don't know
about?

Is it possible to pass a custom php.ini change to a php script via
php-cli? That could be another option if it is possible, but I don't
know enough about it to know if the thought is silly or not.

On Sat, 9 Dec 2023 09:50:44 -0600
Syber Shock <3883@sugar.bug> wrote:

> Is it possible to pass a custom php.ini change to a php script via
> php-cli? That could be another option if it is possible, but I don't
> know enough about it to know if the thought is silly or not.

More brainstorm:

Is the cli invocation of 'php --ini' or 'php -c file.ini -r command'
with custom ini file portable? Could it override disable_function
directives in the server php.ini? That would be like a one or three
liner to solve my particular problem with disabled pcntl_fork.

On Sat, 9 Dec 2023 09:50:44 -0600, Syber Shock wrote:

> On Sat, 9 Dec 2023 13:51:17 +0000 Retro Guy <retroguy@novabbs.org>
> wrote:
>
> <snip>
>
>> Any chance pcntl is disabled?
>
> 100% chance ;)
>
> <snip>
>
> I need to make a workaround. Brainstorm time.

Does the default Bookworm install have this disabled by default?

I don't remember ever needing to enable it in php.ini.

On Sat, 9 Dec 2023 10:02:13 -0600, Syber Shock wrote:

> On Sat, 9 Dec 2023 09:50:44 -0600 Syber Shock <3883@sugar.bug> wrote:
>
>> Is it possible to pass a custom php.ini change to a php script via
>> php-cli? That could be another option if it is possible, but I don't
>> know enough about it to know if the thought is silly or not.
>
> More brainstorm:
>
> Is the cli invocation of 'php --ini' or 'php -c file.ini -r command'
> with custom ini file portable? Could it override disable_function
> directives in the server php.ini? That would be like a one or three
> liner to solve my particular problem with disabled pcntl_fork.

From php manpage:
--php-ini path|file
-c path|file Look for php.ini file in the directory path or use
the specified file

Would that help?

On Sat, 9 Dec 2023 16:08:47 +0000
Retro Guy <retroguy@novabbs.org> wrote:

> On Sat, 9 Dec 2023 10:02:13 -0600, Syber Shock wrote:
>
> > On Sat, 9 Dec 2023 09:50:44 -0600 Syber Shock <3883@sugar.bug>
> > wrote:
> >> Is it possible to pass a custom php.ini change to a php script via
> >> php-cli? That could be another option if it is possible, but I
> >> don't know enough about it to know if the thought is silly or not.
> >>
> >
> > More brainstorm:
> >
> > Is the cli invocation of 'php --ini' or 'php -c file.ini -r command'
> > with custom ini file portable? Could it override disable_function
> > directives in the server php.ini? That would be like a one or three
> > liner to solve my particular problem with disabled pcntl_fork.
>
> From php manpage:
> --php-ini path|file
> -c path|file Look for php.ini file in the directory path or
> use the specified file
>
> Would that help?

See for yourself: https://rocksolid.sybershock.com

I removed all the disable_functions from php.ini for php-cli and
rebooted. This time cron got the nntp daemon running and showing with
PID in procs.

Cron is now pulling data. The local NNTP server is up and lists
newsgroups.

Thanks for the help!