| 1 |
On Sun, Jan 12, 2014 at 11:44:56AM -0600, Dustin C. Hatch wrote: |
| 2 |
> >> I think I found it. It seemed that the integrated run_init support, provided |
| 3 |
> >> through the runscript_selinux.so library that we provide (for OpenRC) didn't |
| 4 |
> >> use PAM authentication, even when policycoreutils was built with USE="pam". |
| 5 |
> >> |
| 6 |
> >> This is because the ebuild didn't use the python-r1.eclass BUILD_DIR |
| 7 |
> >> location (where the files were compiled earlier in the phase) but the |
| 8 |
> >> "normal" ${S} location (which contains the sources). As a result, the "make |
| 9 |
> >> install" phase started building the code, without taking the various USE |
| 10 |
> >> flags into account, and then installing those files. |
| 11 |
[...] |
| 12 |
> > Cool, I've kicked off a catalyst rebuild of my SELinux stage[1234] and |
| 13 |
> > will deploy a new test VM as soon as its done. I'll let you know how it |
| 14 |
> > goes. |
| 15 |
> > |
| 16 |
> > Thanks again for your help. |
| 17 |
> > |
| 18 |
> It is indeed working now, thank you. Is there any chance this can be |
| 19 |
> backported to 2.1? |
| 20 |
|
| 21 |
Certainly; policycoreutils-2.1.14-r5 is now in tree with the same fix. |
| 22 |
|
| 23 |
Wkr, |
| 24 |
Sven Vermeulen |
Archives