Mantis - ISSUES

0000327: Domain settings panel allows setting hard disk space limit to 0MB

A typo means you can set 0MB hard disk space limit for a domain. This causes a divide by zero error on the domain info panel. It should not be possible to set hard disk limit to 0MB.

0000311: Duplicate log entries

Whenever a consistent error occurs like "Client failed to connect because: Address already in use" webcp logs this more than once every minute it seems. Other errors like socket error with webcp-httpd do the same. This makes it a pain to delete the log entries on webcp. It would be much easier if the webcp_log function determined that this error had already happened and ignored it, giving an entry say once every 5 or 10 minutes instead of 25 or more in one minute.

0000116: cron permissions should be configurable

Cron permissions should be configurable - that is, crons should be available to Server, Reseller, Domain and User level users only as given permission by the level above them (permission to create cron jobs and permission to grant the ability to create cron jobs should be seperate). Also, a user should be able to run a script as themselves or as any user which is "under" them (eg. a Domain user could run a script as any one of his Users).

0000326: Problems due to windows carridge returns

Please note any files that have problems due to CRLF from windows systems.

One I recently found:

server/logrotate.inc.phps

This causes logrotate files for domains to be wrongly written as well as for rotating web server logs, which even though not used on most installations breaks logrotate regardless.

0000308: IP Address / sysname fault is not reported.

Web-cp should report when the ip address it knows is wrong, without this the login will always fail if the stored ip address is wrong for whatever reason. Even if webcp's login screen is accessible the ip must be the same for login to work and to use it.

Perhaps if the stored system name is checked against $_Server['ADDR'] then that is enough. The point is that an invalid username or password error appears to override all others, there should be an error for wrong ip because it can prevent login even if the username and password is correct.

0000306: attempts to login to webcp from a browser on localhost fail

I don't know why this is just yet but login attempts from links or lynx as root logged into the virtual machine fail with invalid username or password, this is regardless of the address you are using to access webcp (localhost, 127.0.0.1, 10.0.2.10)

Fails in links and lynx, not able to try any others on localhost..

Logging in at the virtual machine ip in mozilla on the host machine at 10.0.2.10:81 is successful although if the ip webcp-httpd expects (from config.php sysname) is not the same as the address you use to access webcp then login will fail with a message saying that your ip has changed (this is in fact the wrong answer, the problem there is inconsistency between cfg sysname and _SERVER['ADDR'] or whichever server var reveals the ip used to access the site.

These could be related issues. For now it would just be good to find out why login attempts on the local box will not work whereas outside ones will (i've only tested on a virtual machine but that should be perfectly sufficient because it is setup fine) I can test on a physical box but i doubt there will be any difference.

0000325: PHP Configuration errors are ignored.

magic_quotes_gpc must be set to On appears on the log in screen. If we require this setting to be on then the log in screen should not display. Unless the error is non-important, in which case it should not say must.

I believe this affects almost all database queries... normally on seeing it i would just change PHP's setting but if people don't change the setting then databsae queries might not do what they are expected to do.

0000322: Better permissions fallback/handling

This is primarily an installation issue but i think more checking is needed with module setup.

I'll update setup to improve how modules do it and elsewhere in webcp - modules page of server panel..

0000321: SERVER_NAME is not standard

lighttpd and apache seem to give localhost:81 when $_SERVER['SERVER_NAME'] is output, but webcp-httpd gives localhost.. I think this will stop redirects working because the port is added on the end so either we need to filter out the 80 or find a better way to get SERVER_NAME or leave blank and let admins enter it. I noticed webcp-httpd gives out 'er' for $_SERVER['HTTP_HOST'] which may mean its only webcp-httpd which is wrong and perhaps we should use HTTP_HOST instead and just fix webcp's implementation (i think thats a seperate bug report)

0000296: setup fails with webcp-httpd in prefork

Setup seems to fail when webcp-httpd is in prefork mode. See http://www.web-cp.net/forums/viewtopic.php?t=1336 [^] for more details.

0000320: Login screen not shown if there is no skin

The log in screen is never currently displayed unless a skin exists. This is not a problem while we have a default skin. I don't know where noskin could be in effect but if there's no reason for $noskin then it should be removed from the code because it is serving no purpose.

0000319: If running via apache / lighttpd then setup never runs because send_header does not fall back if cfg[httpd_mode] is wrong

We need send_header to fall back on something or else just detect and change httpd_mode if we find it is wrong because httpd_mode affects every send_header().

0000312: Disk usage wrongly reported by server status page

I'm not sure if this affects quotas or anything else, they look to be ok, but on the server info page disk space is reported in kilobytes instead of Megabytes and Megabytes instead of Gigabytes. I don't know if this is to do with the size of the machine's partitions or not yet but i will update this as more info comes in.

0000318: Setup does not test programs to see if they exist.

Setup should not just locate the programs that are missing from $cfg['prog'] it should also make sure that for everything in $cfg['prog'] a file exists on the system. This either means checking each item in the array specified in config.inc.php or config.php to make sure it exists; or making soft links to the files that are required e.g. use ln -s on unix to link /usr/bin/df to /bin/df.

Links could be used to locate service handling programs e.g. /etc/init.d/service could link to /etc/rc.d/service or /etc/init.d/rc.d/service or /etc/rc.d/rc.service or /usr/etc/rc.d/service or wherever the maintainers like to put them.. often /etc/init.d is sufficient.

0000080: multiple language support enhancement

Enhancements I would like to make to the multi-language support:

1. Moved to a separate bug.
2. Make a language-definition tool to enable users to easily create additional translations or update/complete existing translations (this could be part of web-cp or an online tool on the web-cp.net website).
3. Make a guess at the user's language when starting the install process and when first logging in using HTTP_ACCEPT_LANGUAGE and/or something like net_geo.
4. Distribute web-cp with only the most common and/or complete language definitions.
5. Allow direct downloading of additional language modules from the web-cp.net module repository or download with a web-browser.

0000302: skins not compatible with new parser

At least the metroweb and possibly additional skins are not compatible with the new parser code. The templates need to be corrected.

0000301: Interface seems broken (parser not replacing {KEY} with a peice of text)

I think this is to do with the new parsing / caching of lang files. At the log in screen file_put_contents() (web/functions.inc.phps version) was giving an error at line 3906 because it couldn't open the file lang/english.cache.phps.

I found this strange anyway because the lang directory is not writable and no such file existed. So i created one with touch. Same error for web/menu.default.cache.phps

After this the error went away but the login screen now shows {LOGO_MOD}{/LOGO_MOD} and there is no text on the submit button.

Remove the cache files and the submit button has a label but other labels like {LOGO_MOD} still missing (maybe to do with module reigstraiton?) cannot be sure until i have run setup again but there is certainly something not working to do with the cache files.

0000047: statistics packages

modules need to be developed for supporting statistics packages such as awstats or webalizer.

0000315: Domaindata unserialize error

Warning: unserialize() expects parameter 1 to be string, array given in /home/webcp-0.6/webcp/web/functions.inc.phps on line 1029

I'm not sure whats causing this, i can only assume that $domaindata is no longer an array @current($domaindata). I will turn on debugging and see if i can get some more info on this while i turn my attention to backups.

0000171: allow users to backup databases

Users should be able to backup databases that they or their sub-users own now that the owner of a database is enforced to be an actual web-cp user. This was not possible before but with the changes to database ownership in 0.6 this is now possible.

0000313: Mysql result resource errors.

We get errors all over the place from hooks.inc.phps, hdquota.mod.phps, bwquota.mod.phps, and i'm sure there are more out there just waiting to start happening. The problems are with loops that call mysql_fetch_array or mysql_num_rows or similar functions on an expected result resource without checking if it is false, i guess php is more careful these days than it used to be.

This relates to error checking, we often just use if ($error) in webcp without actually checking what is returned, the result depends entirely on PHP's working out of what is false and what is true, which is sometimes not always obvious.

This is not a major issue but if people see mysql warnings all over the place they may think something's wrong when it isn't.

0000317: make languages into modules

Make languages into modules.

0000316: error opening cache file

fopen(lang/english.cache.phps) [function.fopen]: failed to open stream: Permission denied

This appears to happen when the file doesn't exist or when the file exists but is not readable for some reason.

0000314: If the only reseller IP address is changed then domains should be updated

Without this or the admin realising they need to update all domains seperately then everything breaks, i can add a check to get_server_ip to make sure that if no ip addresses match in the NAT mapping that the ip stored is used instead even if NAT is on but this still leaves domains out of date if you have to update the server IP.

Maybe we need some generic checks that webcp should perform whenever the backend does a maintennace operation, say sanity checks or maintennace checks so that the backend can sort out problems. If the backend could fix things itself it would really help admins when things go slightly out of line with what they should be.

0000087: Several interface bugs

* I noticed that it's not possible to edit existing packages. A button next to the "delete package" would be handy. Trivial mistakes would require you to recreate the package (not a big issue, though).

* It might also be handy to be able to assign a package to an already existing domain. This way you can update large amounts of domains easily. Prefferably, this would be both on the "Domain Settings" and the "Domain Management" page.

* The "Domain Management" page doesn't have the same search/edit/delete features as the "User Management" page.

* The "Editable" display of the "User Management" doesn't work intuitivly. It should probably auto check the update box when you make a change. Furthermore, selecting the "Modify Users" option would be nice.

These are some minor issues I found while browsing around the demo. I'm not sure if you can really call them bugs. Nice work so far on the 0.6 version. I whish I could contribute some myself.