[Mailman-Developers] [ mailman-Bugs-596565 ] mailman should use alphabetic U/GIDs

[noreply@sourceforge.net]

Bugs item #596565, was opened at 2002-08-17 17:13
You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100103&aid=596565&group_id=103

Category: configuring/installing
Group: 2.1 beta
>Status: Closed
>Resolution: Accepted
Priority: 5
Submitted By: Todd Vierling (tvierling)
Assigned to: Barry A. Warsaw (bwarsaw)
Summary: mailman should use alphabetic U/GIDs

Initial Comment:
In short: "compiling in numeric IDs is Very Evil."

The long story:

I'm trying to put together a packaging environment for 
NetBSD's "pkgsrc" to allow easy build and install of 
Mailman.  Unfortunately, this is currently impossible 
thanks to the fact that alphabetic IDs are resolved to 
numerics at compile time.  (Nearly all packages 
requiring cutstom IDs resolve them at runtime using get
{pw,gr}nam(3)).

Thanks to this behavior, it is not currently possible to:

* create a prebuilt binary package of Mailman, and 
install it later on a system without the source tarball;
* migrate an existing installation of Mailman to another 
system that has a uid/gid conflict without recompiling.

There are other problems that compiled-in numeric IDs 
cause, but these two really stick out.

I'm not a Pythonhead, so I don't claim to know how to fix 
this correctly.  However, I've attached a diff for 
configure.in that sets the framework for the start of fixing 
this problem.  In particular, the diff changes the 
MM_FIND_*_ID macros to report the ID that matched 
without resolving names to numbers, and allows the 
build to proceed if forced --with-{user,group}name 
options are specified and those IDs don't exist at build 
time.  (The latter case helps autopackaging in systems 
like NetBSD's pkgsrc, where the IDs may be 
autocreated *between* the "configure" and "make 
install" phases.)

The things that still need to be done include fixing the 
main scripts, and the checkperms logic, to cope with 
either numeric or alphabetic IDs; and possibly adding a 
runtime configuration of IDs (which really helps if the 
admin needs to reconfigure or change the mail or Web 
servers, while avoiding to rebuild Mailman).

I openly welcome Python gurus to look into what's 
*really* needed to fix this....


----------------------------------------------------------------------

>Comment By: Barry A. Warsaw (bwarsaw)
Date: 2002-08-23 16:03

Message:
Logged In: YES 
user_id=12800

I've thought about it more and I am going to apply this patch.

----------------------------------------------------------------------

Comment By: Barry A. Warsaw (bwarsaw)
Date: 2002-08-22 18:45

Message:
Logged In: YES 
user_id=12800

I like this idea a lot, even though the patch is far too
incomplete.  Fortunately, you motivated me to hack the code
to support symbolic names instead of numeric ids.

Attached is a patch I intend to commit for MM2.1b4.  Would
you please try it and see if it accomplishes what you need?
 One minor ugliness is that it retains --with-cgi-gid and
--with-mail-gid as the configure switches to set the wrapper
group names.  Given the communities knowledge about these
switches, I'm inclined not to change or deprecate them.

----------------------------------------------------------------------

You can respond by visiting: 
https://sourceforge.net/tracker/?func=detail&atid=100103&aid=596565&group_id=103