I want to hear from downstreams that want to add symbol-versioning to
FreeType first, why do they need to.

On Tue, Jan 30, 2018 at 12:22 PM, Tom Kacvinsky <tkacv...@gmail.com> wrote:

> You raise a fair point.  The point I was trying to make, and perhaps I
> didn't make it clear enough
> what I was after, is this: suppose some downstream maintainer decides to
> add symbol versions.
> Wouldn't it be nice if we already had that so there is no mess between how
> one Linux distribution
> does it versus another?  Granted, there still might be a mess, but we can
> mitigate the hassle.
>
> And, now for another idiom:  that's just my two cents.
>
> Regards,
>
> Tom
>
> On Tue, Jan 30, 2018 at 3:03 PM, Behdad Esfahbod <beh...@cs.toronto.edu>
> wrote:
>
>> I know what symbol versioning is used for (I was also package maintainer
>> at Red Hat for four years). But don't see how it applies to FreeType.
>> FreeType never changes ABI in backward-incompatible way. Its build system
>> is already adhoc enough. I don't want to see more complexity added
>> unnecessarily. Also, I don't want FreeType binaries to come with the same
>> headaches that libpng and openssl cause every time they bump soname.
>>
>> On Tue, Jan 30, 2018 at 4:45 AM, Tom Kacvinsky <tkacv...@gmail.com>
>> wrote:
>>
>>> Hi all,
>>>
>>> I'll see what I can do.  To be honest, the only platforms that really
>>> support this
>>> are Linux and Solaris.  I definitely have access to Linux machines, but
>>> not a
>>> Solaris machine.  I might be able to get access to the latter.
>>>
>>> Despite all of the talk about whether symbol versioning is useful (and
>>> this is not
>>> meant to be snarky), keep in mind the major commercial Linux
>>> distributions use
>>> symbol versioning, as well as the free Linux distributions.  I work for
>>> SUSE and
>>> my colleagues highly recommend getting symbol versioning into FreeType.
>>> I
>>> agree with them and am willing to do the work as I find time -  as the
>>> saying goes,
>>> put my money where my mouth is.
>>>
>>> The only thing I would need is a way of getting the API functions to add
>>> to a
>>> symbol versioning linker file.  This way, the symbol versioning script
>>> doesn't
>>> need to be checked in, and if we add something to the API, we don't need
>>> to
>>> worry about missing it in the library version file.
>>>
>>> Regards,
>>>
>>> Tom
>>>
>>> On Tue, Jan 30, 2018 at 1:25 AM, Werner LEMBERG <w...@gnu.org> wrote:
>>>
>>>>
>>>> > Admittedly, symbol versioning is very poorly supported and
>>>> > documented [...]
>>>>
>>>> Yes.
>>>>
>>>> > [...] All in all, this looks like something to stay away from.
>>>>
>>>> Maybe there are more knowledgeable people who want to contribute...
>>>>
>>>>
>>>>     Werner
>>>>
>>>
>>>
>>> _______________________________________________
>>> Freetype-devel mailing list
>>> Freetype-devel@nongnu.org
>>> https://lists.nongnu.org/mailman/listinfo/freetype-devel
>>>
>>>
>>
>>
>> --
>> behdad
>> http://behdad.org/
>>
>
>


-- 
behdad
http://behdad.org/
_______________________________________________
Freetype-devel mailing list
Freetype-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/freetype-devel

Reply via email to