[Development] Qt 5 and old versions of XCB

[Craig.Scott at csiro.au]

On 18/03/2012, at 11:53 AM, Thiago Macieira wrote:

> On domingo, 18 de março de 2012 11.43.02, Craig.Scott at csiro.au wrote:
>> That's fine when you are only building an application to run on your
>> machine, but for people who want to distribute Qt5 apps, this would be an
>> undesirable solution. You'd end up having to ship the xcb libraries as well
>> as those libraries xcb depends on as part of your package.
> 
> I'm sorry, but we can't keep supporting very old systems. By the time of the 
> release, OpenSUSE 11.1 will be 18 months old or older.

Wow..... are you suggesting that we're only interested in ensuring Qt5 runs on systems that 18 months old or newer? That's going to make a lot of enemies.....


> 
>> I asked about xcb on the LSB mailing list and early responses suggest that
>> although xcb has been mentioned there before, it doesn't sound like it's
>> progressed much beyond that. Dropping support for xlib and relying on xcb
>> would thus make Qt5 a non-starter for some people.
> 
> No one is working on xlib and it's not a reference platform. XCB is and it's 
> been around for years.
> 

Not arguing with you on this one. But my question is how consistent is xcb support across the major linux distributions? If ISV's can't build a binary that they can reliably distribute, that's going to be a serious road block for them to move to Qt5.


> The fact that the LSB is stuck in time is not a reason to hold Qt back from 
> important technology. I think it's important we do the opposite: help LSB 5.0.


The original question for this thread was about whether xlib should be dropped. As others have indicated, there are still plenty of people who rely on / use xlib over xcb. If some people are still finding xlib useful and if there are cases where xcb support might not be an option, is it really wise to drop xlib? Regarding LSB, this discussion here is part of the motivation for the LSB guys to look more closely at xcb (which is a good thing), but simply writing off the LSB because it is inconvenient for you is a bit heavy handed. I can't overstate the usefulness the LSB has been to us with the Qt4 series. If Qt5 didn't build with the LSB, we would stick with Qt4 until the LSB caught up to wherever Qt5 moved on to. Not saying everyone else would do the same, but if you aren't being forced to create that situation by dropping xlib, why do it?


--
Dr Craig Scott
Computational Software Engineering Team Leader, CSIRO (CMIS)
Melbourne, Australia