[Development] Removal of xlib plugin - a possible disaster.

[Rick Stockton]

On 03/18/2012 03:23 AM, Robin Burchell wrote:
> On Sun, Mar 18, 2012 at 7:59 AM, Rick Stockton
> <rickstockton at reno-computerhelp.com>  wrote:
<< SNIP >>
>> I can't think of a WORSE spectacle, for the reputation of Qt-Project,
>> than the Release of an Alpha Build in which large numbers of exerpienced
>> Qt users on Linux Linux _might_ be unable to use the XCB plugin and show
>> a main Window, before getting some kind of 'ABEND', with _any_ of our
>> own provided examples.
> I agree: If we are shipping an alpha in which the xcb plugin is
> unusable for large numbers of people (though this needs to be
> quantified - how many people are hitting these problems?), then we're
> doing it wrong. We shouldn't ship something that doesn't work.
>
>> With these two bugs present (i.e. SIGSEGV happening on some "older"
>> Distros, and "Undefined Symbol" happening on RPMs built just 6 days
>> ago), you NEED to have an alternative, EASY plugin for these people to
>> use. Even at alpha.
> No. As I said, the problems need to be *fixed*. Providing workarounds
> just means that people that can _find_ the workarounds keep using
> those workarounds, those that _can't_ find them think "Qt 5 is crap,
> it doesn't work", and when that workaround is under/unmaintained code,
> that won't leave a good impression anyway.
>
>> (B) stop ignoring the IRC "I'm getting
>> 'segment fault' with an example program" postings
> IRC is *not* a bugtracker. It's transient. You're shouting into a
> void, and if someone who happens to be able to help you reads what
> you're saying, and has some time to help you - great. That doesn't
> mean you should treat it as the one place to report issues. Writing to
> the mailing list is a more sensible idea, as it has a much more
> permanent record. Filing bugs (or even better: patches), and bringing
> them to the attention of the people who work on that code (through
> mailing lists, or if you _know_ who to speak to, IRC) is even better.
>
> Relatedly, quoting you from Gerrit:
>
>> I have been present when OTHERS have brought up the same problem (crash with
>> SIGSEGV, before bringing up the main window) on OTHER Distros, starting sometime in mid (or late?) February.
>>
>> No one addressed it, and I've got no experience in this area. [w00t has been logged
>> in at the time of least 2-3 of those queries... but possibly in unannounced "back later"
>> mode at those times.]
> Unless something goes wrong with my server, I'm always "connected",
> but not necessarily reading, or even in front of the computer. I
> connect to IRC through a server, and then connect to _it_ when I'm
> around. Sometimes, I might read things that happened while I was away,
> but not often.....
Exactly a point which I attempted to make. But I didn't emphasize that 
you were ALMOST CERTAINLY, because my word "possibly" is not a good word 
for "almost certainly". Thanks for making this more clear than I wrote it.

Bottom line: everyne who hears of such an event on Qt5, on any of our 
IRC channels, should _beg_ the person to open a QTBUG for it.
>   I have two seperate such connections (one for work, one
> for "personal" stuff), and between them, I'm on around 70-80 different
> channels. If I tried to keep track of everything that happened in
> those channels all of the time, I'd never get anything else done. So I
> don't. I know that many other people use IRC in similar ways....
Your 100% right, of course. I myself am Away, without being marked 
"Nick_AWAY", frequently.
>> and POSTPONE ALPHA until we've got them fixed.
> I think we'd need to really have an idea of what the problem is, how
> many people it affects, and the effort to fix it, before we can make
> that call. But yes, as I've said: I don't think that shipping broken
> code is a good idea.
Strongly agreed, and I think that we understand each other. We just 
recommend different approaches: I see the Alpha as a good tool to get 
this bug quantified AND possibly more isolated, well-defined -- by 
having other people (with a far wider range of Linux system 
configurations) experiencing cases "in the real world". There is a 
significant probability that it WON'T be happen to a lot of "normal 
users": my own system had a strange combination of old xcb core and xcb 
utils, but with new Wayland/Mesa/Cairo etc.

I'd like to see Alpha-1 cut sooner, rather than later (even if it 
contains severe bugs, with workarounds such as xlib plugin). You'd 
prefer higher quality, putting "XCB fails to initialize a main Window 
with size hints" on the list of Releaes-Critical bugs, even for an Alpha 
Release. We're in agreement that the one thing NOT to do is put out 
alpha _with_ show-stopper XCB bugs present, *AND* xlib plugin removed. 
You recommend to fix the show-stopper bugs first, I recommend to leave 
xlib in as a workaround.

Thanks for a great discussion! I will not be changing my review vote (on 
removing Xlib) before XCB again becomes usable on my own system. Dr. 
Scott, and others, have brought up other issues of great interest -- but 
the first choice, probably, is whether or not to hold up Alpha while 
bug-hunting. Not up to me, I only recommend. :))