1 |
On 11/30/2016 11:45 AM, Mike Gilbert wrote: |
2 |
> On Wed, Nov 30, 2016 at 11:28 AM, Michael Orlitzky <mjo@g.o> wrote: |
3 |
>> On 11/26/2016 12:47 PM, Mike Gilbert wrote: |
4 |
>>> The values get clobbered immediately afterward, so why bother? |
5 |
>>> ... |
6 |
>>> qeinfo "Determining the location of the kernel source code" |
7 |
>>> - [ -h "${KERNEL_DIR}" ] && KV_DIR="$(readlink -f ${KERNEL_DIR})" |
8 |
>>> [ -d "${KERNEL_DIR}" ] && KV_DIR="${KERNEL_DIR}" |
9 |
>>> |
10 |
>> |
11 |
>> This changes the behavior if $KERNEL_DIR is a symbolic link to a |
12 |
>> nonexistent directory, doesn't it? |
13 |
> |
14 |
> Yes, I suppose it does. Do you anticipate that will cause a problem? |
15 |
> |
16 |
|
17 |
*shrug* |
18 |
|
19 |
There's a bug there, but who knows what the code was supposed to do. The |
20 |
docs say that KV_DIR is |
21 |
|
22 |
...a string containing the kernel source directory, will be null if |
23 |
KERNEL_DIR is invalid |
24 |
|
25 |
To me, that makes the "readlink" line the bug, because KV_DIR will be |
26 |
set even if KERNEL_DIR points nowhere. But then why is the readlink |
27 |
there? Should KERNEL_DIR support symlinks? (Probably, yes.) If so, the |
28 |
second test should be something like |
29 |
|
30 |
[ ! -d "${KV_DIR}" ] && KV_DIR="" |
31 |
|
32 |
to undo the previous line when KERNEL_DIR didn't lead anywhere. A better |
33 |
fix would avoid setting KV_DIR entirely in that case. |