Re: [Libofx-devel] libofx symbol visibility patch

[Bryan D. <bd...@gm...>]

2011/6/7 Benoit Grégoire <be...@co...>:
> On June 4, 2011 11:06:55 PM Bryan Donlan wrote:
>> Hi,
>>
>> As mentioned some time before, I've wrote a patch for libofx[1] to
>> hide internal symbols. By hiding internal symbols, we can avoid
>> external programs inadverently using internal (subject to change)
>> symbols by making it very explicit what is considered internal or not.
>> This has already turned up one case of poorly documented ABI symbols
>> (ofx_*_msg), which I've also added to the header. Additional
>> information can be found at the comment block at the top of the patch.
>
> Applied
>
>> Because this does change the ABI slightly by deleting symbols (alebit
>> only ones not listed in the public header), it may be safest to bump
>> the soname if this patch is applied. Note I don't intend to apply this
>> patch to the debian packaging without upstream blessings, as it
>> changes the ABI in this manner.
>
> Ok, playing it safe.
>
>> Also, you had mentioned you'd like to roll my changes into a 0.9.5 to
>> keep everything in sync - are you likely to release soon? I have some
>> fixes for the debian package itself that I'd like to upload soon, but
>> if 0.9.5 is coming out soon I'll wait for it.
>
> Well, I think we'll be pretty much ready for a 0.9.5 when you are.

I'm pretty much ready whenever an upstream release goes out - just
have to remove my debian patches, so the usual soname change
adjustments, test the result, and send it upstream (and then
coordinate binNMUs afterward for reverse-dependencies, of course).
Just let me know :)

Thanks,

Bryan