[prev in list] [next in list] [prev in thread] [next in thread] 

List:       kde-core-devel
Subject:    Re: the importance of -ansi
From:       "Jos van den Oever" <jvdoever () gmail ! com>
Date:       2007-06-16 8:51:44
Message-ID: c2dbc4260706160151y7dfc5076q899f8071f5cd1476 () mail ! gmail ! com
[Download RAW message or body]

2007/6/16, Thiago Macieira <thiago@kde.org>:
> I thought it was a stale build because that had happened to me once. I was
> getting crashes I could not understand and valgrind was reporting memory
> violations ("address is 0 bytes after a block of XX bytes at 0xHHHHHH")
> that shouldn't be there. That's exactly the same symptom Jos was having.
>
> My problem was fixed by removing all .o and rebuilding.
>
> But when I started reading CLucene code, I was thinking "whoever released
> this should be sent back to school". Sorry, but the code is nasty. It
> changes class sizes according to macros like _DEBUG. That's just asking
> for trouble.
>
> That is still my opinion. CLucene needs to be cleaned up so that it does
> NOT depend on macros.
>
> In any case, I'd never have guessed math.h does the same thing. Then
> again, I'd never heard about float_t. Why is that being used at all? It's
> nowhere in glibc's TeXinfo documentation.
CLucene will never win a beauty pageant, that's for sure. The code is
remarkably stable though (when built right). The reason for that is
that it is used a lot. CLucene is so nice because it is really fast.
Lucene 2.1 is a nice point for starting on a clean new CLucene
version, but good luck finding somebody with the time to implement it.

Currently CLucene is the best backend I know, but if a better one
comes around we can switch relatively easily. A backend requires 3
classes to be implemented. Right now we are expanding our general
backend unit tests to ensure quality and fast development of all
backends.

Cheers,
Jos
[prev in list] [next in list] [prev in thread] [next in thread]