1 |
W dniu sob, 10.03.2018 o godzinie 01∶05 +0000, użytkownik Robin H. |
2 |
Johnson napisał: |
3 |
> On Thu, Mar 08, 2018 at 06:05:43PM +0100, Michał Górny wrote: |
4 |
> > Make linux-info_get_any_version die if it can't determine any version |
5 |
> > of the Linux kernel. This indicates a problem with the eclass code |
6 |
> > (as it should not happen on Linux) and the missing KV_* variables |
7 |
> > are going to cause random misbehavior and failures. |
8 |
> |
9 |
> This change worries me a bit. get_running_version calls get_version in |
10 |
> some cases, and there are corner cases there which could trip this up. |
11 |
> |
12 |
> linux-info_get_any_version -> get_running_version -> get_version |
13 |
> with get_version returning non-zero. |
14 |
> |
15 |
> get_version has ALREADY returned non-zero once in that path, and we're |
16 |
> potentially tweaking KERNEL_DIR/KBUILD_OUTPUT before calling it again. |
17 |
> |
18 |
> Most of the breakage cases IIRC are where |
19 |
> /lib/modules/${KV_FULL}/*/Makefile still exist, but point to kernel |
20 |
> sources in weird places that are broken. |
21 |
|
22 |
Yes, and this needs to be fixed one way or another. Otherwise you don't |
23 |
get KV_* exported and a lot of other logic trips on unset variables. |
24 |
|
25 |
-- |
26 |
Best regards, |
27 |
Michał Górny |