[Interest] [#ID:INC-1251018#] Installation issue.

Thiago Macieira thiago.macieira at intel.com
Tue Mar 12 16:45:20 CET 2019


On Tuesday, 12 March 2019 02:10:08 PDT Dr.-Ing. Christoph Cullmann wrote:
> > But the point remains: at some point, it becomes difficult to actually
> > build those dependencies because the distribution is old. Also note Qt 6
> > will not ship bundled libraries. There will be a way to install them from
> > sources, but then we go back to the problem of actually building them.
> 
> I think this will be a hard issue for a lot of people that need to support
> industrial customers.

Understood. We do mean to provide a way to perform an automated build of those 
dependencies, both as static and as dynamic libraries, so you should get the 
same benefits as current bundling. I was going to write that I didn't expect 
it to be well tested on Linux, but then I realised that the binaries from 
qt.io are likely to use this technique, so I withdrew my argument before even 
sending it.

Advantage of this method is that we always get the latest, with latest 
security fixes. If there are futher fixes that apply to users of Qt, then all 
they need to do is rebuild. They will see clearly what those libraries are. 
And it simplifies our own maintenance, of course.

But yes, a disadvantage is now having to deal with the buildsystem for those 
libraries.

> AbsInt will migrate to Red Hat/CentOS 7 for our builts to circumvent that,
> but I think a lot of other people don't have this possibility.

RHEL 7 was initially released in 2014. If you can't update to a 6-year-old 
series in 2020 when upgrading a major Qt, you could have much bigger problems.

I'd also advise looking into RHEL 8, which should be released this year, for 
further long-term support.

> For the xkbcommon thing: What me most disturbed is that it was removed
> during a patch release. 5.12 did compile fine on CentOS 6.x, 5.12 did have
> it removed and failed. That was kind of unexpected.

Yeah, not ideal. But we had to do it because of 5.12's long term nature. It 
ought to have been done for .0, but we just couldn't in time.

Rock, meet hard place.
-- 
Thiago Macieira - thiago.macieira (AT) intel.com
  Software Architect - Intel System Software Products






More information about the Interest mailing list