| 1 |
maillog: 19/06/2006-17:15:37(-0700): Robin H. Johnson types |
| 2 |
> On Mon, Jun 19, 2006 at 05:00:41PM -0700, infowolfe wrote: |
| 3 |
> > >Kernel headers being the virtual/linux-headers dependency that Georgi |
| 4 |
> > >mentioned. `uname -r` works, but is annoying because you can't build |
| 5 |
> > >for a kernel other than the one you're running. |
| 6 |
> > Which only applies to kernel modules, not things like gnupg that don't |
| 7 |
> > REALLY need kernel sources in order to function. |
| 8 |
> Gnupg builds it's secure memory functionality differently based on what |
| 9 |
> is available from the kernel. All of the possible APIs are available in |
| 10 |
> the headers, but depending on what the kernel is configured as, affects |
| 11 |
> which of the APIs provide secure memory blocks. |
| 12 |
> |
| 13 |
> With GnuPG, it happens that on older LiveCDs, the kernel that is running |
| 14 |
> from the LiveCD doesn't offer what it wants, but the one that you would |
| 15 |
> be rebooting to does. |
| 16 |
> |
| 17 |
> Could upstream have handled it better? Yes, most definitely. Did they? |
| 18 |
> No, not yet. We're stuck picking up the pieces. |
| 19 |
|
| 20 |
What does upstream have to do with the decision to "chmod u+s,go-r |
| 21 |
/usr/bin/gpg" or not? |
| 22 |
|
| 23 |
-- |
| 24 |
(* Georgi Georgiev (* YAAH! DEATH TO OATMEAL! -- Calvin (* |
| 25 |
*) chutz@×××.net *) *) |
| 26 |
(* http://www.gg3.net/ (* (* |
Archives