1 |
On 10/27/10 1:05 PM, Diego Elio Pettenò wrote: |
2 |
> Il giorno mer, 27/10/2010 alle 12.49 +0200, "Paweł Hajdan, Jr." ha |
3 |
> scritto: |
4 |
>> Some context first. Upstream does not promise any binary compatibility |
5 |
>> between releases, so each version has a different SONAME, if any. |
6 |
> |
7 |
> That's a good approach (although a bothersome for its users).. |
8 |
|
9 |
By the way, what do you think about |
10 |
<http://ispras.linux-foundation.org/index.php/ABI_compliance_checker> ? |
11 |
|
12 |
>> Currently dev-lang/v8 does not use SONAME (I think that's the default), |
13 |
>> because I'm not sure what's the best way to handle it. |
14 |
> |
15 |
> That's bad, it _has_ to have a SONAME, otherwise you're indicating that |
16 |
> al the versions have the same ABI (which as you just said is not the |
17 |
> case). |
18 |
|
19 |
That's one of the reasons it's still hard masked. |
20 |
|
21 |
> Okay now you're confusing the two concepts. The DT_SONAME tag has |
22 |
> nothing to do with the name on the filesystem for what concerns the link |
23 |
> editor, it's used by the dynamic loader. |
24 |
> |
25 |
> The link above would be fine though, just let it link as libv8.so, and |
26 |
> then be loaded as libv8-1.2.3.so. |
27 |
|
28 |
Ah, thanks! I think I understand it much better now. Bumped v8, now with |
29 |
SONAME and a symlink. I was able to compile www-client/chromium against |
30 |
that, and added an experimental system-v8 USE flag to it. |