umask package policy

[Warren Togami]

Ville Skyttä wrote:
> On Sun, 2005-05-22 at 12:13 -1000, Warren Togami wrote:
> 
> 
>>Should we make it a packaging policy that packages must own all 
>>directories and files that it installs in order to avoid umask 077 
>>problems like this where the installed software is effectively broken?
> 
> 
> +1, although I thought that already is a policy at least in Extras.  But
> not _all_ directories it installs, only those that are not owned by its
> prerequisite packages.
> 

Hmm you are right, but that makes it more difficult to make an automated 
test.  I guess both tests will require a chroot and installing all deps.

> 
>>Also the latter bug is %post creates files with umask 077, leading to 
>>similar breakage.  Should then the rule be "use umask within %post if it 
>>creates files?"
> 
> 
> I think it would be simpler and cleaner to have rpm execute all
> scriptlets with let's say a 022 umask, and make that umask value
> configurable in eg. /usr/lib/rpm(/$vendor)/macros.
> 

This isn't a solution for older distributions, this umask problem is 
troubling for RHEL too, and we can't push this rpm change there.

> 
>>Can rpmdiff check for this in the future?
> 
> 
> Are you referring to the rpmdiff distributed with rpmlint, or something
> else?

Automated tests in general.

Warren Togami
wtogami at redhat.com