[Qt-creator] Creator stability

[Bill King]

On 03/26/2010 05:06 PM, Raggi Roberto (Nokia-D-Qt/Berlin) wrote:
>
> Hi,
>
> On Mar 26, 2010, at 12:15 AM, ext Bill King wrote:
>>>
>> In this case I got a "doesn't happen here" and then it was dropped
>> (maybe it was worked on internally), but the email responses from
>> roberto showed that it had pretty much been ignored (even after me
>> adding extra information, and then even a patch to fix the issue).
>
> I'm pretty sure the only email I sent about this topic was a short
> reply on our internal mailing list where I just said 1) we can't
> reproduce it 2) we don't have a working valgrind for OSX 10.6 and 3)
> the email ends with - I guess we have to investigate what's going
> wrong with the binder.
The original bug report was not transitioned from reported to accepted
for over 3 weeks. Considering our recent focus on doing at the very
least that, yeah, that's the issue I was trying to point out.
>
> We applied your patch (commit b0b95f88756dbdf4981c97a325734300a65d8268
> ) but unfortunately it was creating quite a number of regressions in
> our refactoring engine so we had to revert it
> (change 933e52888e8e69f876bdf4640379ecdddb2c2c9c ).
>
> And by the way, we tried to reproduce the crash on 4 different
> computers. We tried Linux, OSX and Windows and we also checked with
> memcheck so who knows, maybe we can get some help from the guys on
> this mailing list. Please, can someone try to reproduce this
> bug http://bugreports.qt.nokia.com/browse/QTCREATORBUG-807 ? I'm sure
> we will have a better understanding of the problem if we have more
> reports.
I had the crash on linux and 2 windows boxes, and both purify and
memcheck reported the issue. Very odd.
>
> About the quality of the C++ front-end. Well, i can tell you that we
> are testing it quite a lot. 
> We have 3 tools that we use daily to test the front-end with the gcc
> test suite and with the Qt source code (in creator/tests/manual,
> cplusplus-dump, cplusplus, plain-c++).
Testing's good :D
>
> Yes, it can happen that we introduce regressions but that's because an
> incremental front-end for C++ need to parse both correct and
> invalid(unfinished?) code and i'm sure you can see how many "wrong"
> configuration you can have when parsing a few million lines of C++
> code so the test suite needs to be a bit special and that's the reason
> why we're using gcc/gcc/testsuite.
>
> I think your attack to us is wrong and not fare.
Fair/not fair, at the end of the day, we're professionals, and as such,
we have to take the good with the bad. Part of that is maintaining
communication, part of it also is realising when things are going wrong
and buckling down to rectify that situation. I've been talking with Ed,
I'll leave it with that for now :)
>
> ciao robe
>


-- 
Bill King, Software Engineer
Qt Development Frameworks, Nokia Pty Ltd
Brisbane Office