On 08/24/2017 08:19 AM, Adam Williamson wrote:
On Thu, 2017-08-24 at 09:12 +0200, Remi Collet wrote:
> Le 24/08/2017 à 08:58, Adam Williamson a écrit :
>> On Thu, 2017-08-24 at 06:44 +0200, Remi Collet wrote:
>>> Le 24/08/2017 à 02:32, Moez Roy a écrit :
>>>
>>>> The soname bump requires packages that depend on it need to be rebuilt,
so
>>>> I updated ImageMagick to 7.0.6-9.
>>>
>>>
>>> Such a version bump in stable branch is not acceptable.
>>>
>>> ImageMagick 7 removed lot of functions (deprecated in v6)
>>>
>>> Especially as ImageMagick 6 is still maintained.
>>>
>>> BTW, latest version 6.9.9-9 also have some API changes
>>>
>>> - 6.9.6-8 => bump to .3
>>> - 6.9.7-6 => bump to .4
>>> - 6.9.9-0 => bump to .5
>>>
>>> Yes, this package is a nightmare.
>>
>> Note the versioning is a bit subtle. What was released upstream was
>> 6.9.9-9 , which in our versioning would be 6.9.9.9 . Upstream *didn't*
>> go from 6.9.8 to 6.9.9, but from 6.9.9-8 to 6.9.9-9. The package in
>> Fedora before all this mess was versioned 6.9.9.3 , which I believe
>> would be upstream's 6.9.9-3 , and I *think* there is no ABI/API
>> incompatibility between 6.9.9-3 and 6.9.9-9. The soname in the 6.9.9.3
>> packages was already
>> "libMagickCore-6.Q16.so.5()(64bit)".
>
> Yes in F27+ (6.9.9.3)
>
> F25/F26 have 6.9.3.0 (upstream 6.9.3-0) with
> libMagickCore-6.Q16.so.2
> libMagickWand-6.Q16.so.2
Yep, I just looked back and saw that nirik built upstream 6.9.9-3 but
hadn't actually sent it as an update, since it was an soname bump and
all the necessary rebuilds weren't done yet :/
Sigh. Yeah, I didn't want to push 7.x to even rawhide without
investigation that all dependent packages were ready to move to it.
I was going to look at f25/f26 for a 6.x update, but I haven't had any
time to get to it. ;( Would need a bunch of rebuilds and waiting in
testing a while so 3rd parties could see it and rebuild at the very least.
kevin