| 1 |
On 02/25/2015 03:41 PM, Anthony G. Basile wrote: |
| 2 |
> On 02/25/15 15:38, Zac Medico wrote: |
| 3 |
>> On 02/25/2015 12:01 PM, Anthony G. Basile wrote: |
| 4 |
>>> On 02/25/15 14:51, Anthony G. Basile wrote: |
| 5 |
>>>> On 02/22/15 01:30, Zac Medico wrote: |
| 6 |
>>>>> On 02/21/2015 10:22 PM, Zac Medico wrote: |
| 7 |
>>>>>> If we put the real/canonical libstdc++.so path in the DT_NEEDED |
| 8 |
>>>>>> section, |
| 9 |
>>>>>> then it will automatically work with existing soname dependency |
| 10 |
>>>>>> support. |
| 11 |
>>>>> |
| 12 |
>>>>> Actually, we'd also have to add a way for you to put the full path of |
| 13 |
>>>>> the libstdc++.so in PROVIDES. For example: |
| 14 |
>>>>> |
| 15 |
>>>>> PROVIDES_ABSOLUTE="/usr/lib/gcc/*/*/libstdc++.so.6" |
| 16 |
>>>>> |
| 17 |
>>>> |
| 18 |
>>>> I guess I don't understand how this would work exactly. What if |
| 19 |
>>>> someone |
| 20 |
>>>> has gcc-4.8.3. Builds library libfoo.so which uses c++. Then upgrades |
| 21 |
>>>> to gcc-4.9, removes 4.8 and then tries to build bar which is also |
| 22 |
>>>> written in c++ and links against libfoo.so. We would have mismatching |
| 23 |
>>>> abis. How would this catch it and trigger the correct rebuilds? |
| 24 |
>>>> |
| 25 |
>>>> Unless I'm misunderstanding your *'s in that line. Are you using |
| 26 |
>>>> PROVIDES_ABSOLUTE as a way of recording what version of the compiler |
| 27 |
>>>> libfoo.so was build with? So that you'd have a line that says |
| 28 |
>>>> libfoo.so |
| 29 |
>>>> links against /usr/lib/gcc/x86_64-pc-linux-gnu/4.8.3/libstdc++.so, so |
| 30 |
>>>> that parsing that line gives 4.8.3? |
| 31 |
>>>> |
| 32 |
>>>> Also if you had the absolute path in VDB somewhere, like in PROVIDES, |
| 33 |
>>>> then you don't need it in the elf's rpath which would make me feel |
| 34 |
>>>> better. |
| 35 |
>>>> |
| 36 |
>>> |
| 37 |
>>> Actually no, you'd still need rpath for the elf itslef otherwise you can |
| 38 |
>>> still link against the wrong version of libstdc++.so. Note in my |
| 39 |
>>> following example that even though I build test.cpp with 4.7.3 I still |
| 40 |
>>> wind up linking aginast 4.8.3. |
| 41 |
>> |
| 42 |
>> If DT_NEEDED contains the absolute libstdc++.so path, it's guaranteed to |
| 43 |
>> link against the correct version, regardless of rpath. |
| 44 |
>> |
| 45 |
> |
| 46 |
> How do you get DT_NEEDED to the absolute libstdc++.so path when building? |
| 47 |
|
| 48 |
I'm not sure how we're going to accomplish that yet, but it should be |
| 49 |
feasible. Ideally, the build system would support it. The worst case |
| 50 |
would be that we would have to patch the DT_NEEDED sections after |
| 51 |
src_install. |
| 52 |
-- |
| 53 |
Thanks, |
| 54 |
Zac |
Archives