1 |
On 30 Mar 2010, at 03:13, Fabian Groffen wrote: |
2 |
|
3 |
> On 30-03-2010 07:19:02 +0200, Markus Duft wrote: |
4 |
>> just to chime in on the movability: i successfully moved whole prefix |
5 |
>> installations to other directories, and they worked fine (hint: i used |
6 |
>> chpathtool to adjust paths) the only limitation is, that the path has to |
7 |
>> be the exact same length (or less long, but that one didn't work for me |
8 |
>> yet, but that could be fixable). of course this is not really |
9 |
>> convenient, but it should be possibly for you to create |
10 |
>> |
11 |
>> "/home/user/prefix-with-long-path1" |
12 |
>> and |
13 |
>> "/cluster/disk25/user/prefix-short" |
14 |
>> |
15 |
>> those would be exchangeable. Of course chpathtool is possibly still not |
16 |
>> perfect, and i observed the one or the other segfault with certain files |
17 |
>> (none of those lately though). However you're free to fix it ;) |
18 |
> |
19 |
> That could match chpathtool's changes that it got recently. Instead of |
20 |
> trying to shorten paths, it now pads them with leading slashes, e.g.: |
21 |
> |
22 |
> original prefix: /path/to/my/long/prefix |
23 |
> target prefix: ////////my/short/prefix |
24 |
> |
25 |
> As a result, the target prefix is always as long as the original prefix, |
26 |
> be it somewhat ugly. We found that this solves misc issues that you get |
27 |
> when you suddenly make a string smaller ;) |
28 |
|
29 |
Apologies for resurrecting this topic, and I just realised I've probably been confusing chpathtool with chrpath during all previous discussions, but I haven't seen any mention of patchelf. I got the Intel compiler running in a Linux-prefixed-portage last week. |
30 |
|
31 |
// Cheers; Johan |