[Gambas-user] Re: "[12] Not an object" error.

[Jesper Holm]

On Monday 08 March 2004 18:59, ron wrote:
> On Monday 08 March 2004 15:19, Benoit Minisini wrote:
> > On Monday 08 March 2004 14:35, ron wrote:
> > > As Benoit already send to me as answer
> > >
> > > You must be root to launch 'gbi -a', as this program will write the
> > > info files in the /opt/gambas/lib/info directory.
> > >
> > > The -a options tells gbi to make all info files.
>
> sniper snap
>
> > "gbi -a" write a *.info file for each *.component file it finds
> > in /opt/gambas/lib. In a correct install, if a component is disabled, the
> > corresponding *.component file is not installed.
> >
> > If the ./configure script disables a component, and after all the info
> > file is installed, then this is a bug and you should tell me!
>
> When I am right this is not the case! He wrote the next sentence.
>
> >> When using the Gentoo ebuild, which effectively just removes the gbi
> >> part of the Makefile, I get the below mentioned errors.
>
> It looks to me that gentoo has done the damage, they removed the gbi part
> in the make part to where the .info files are created if I understand him
> right.
>

Thanks Ron, The problem is that the source when compiled "from scratch" so to 
speak, segfaults in the 4th gbi file. All Gentoo did was to comment out the 
generation of .info files in the Makefile.

I don't know whaht these .info files contain, do or if they are vital, but the 
program crashes without them being installed. with "[12] ..." error.


> I give only the way how to do it himself. For him to verify all is done ok
> i give the list what and where he must find it.
>
> In case of a error in one of the library sources for this operation gbi
> abort total. I had it in the past after the third file and only 1 and 2
> where written, 3 was ok but not written and the fault was in the 4'th. The
> rest was not done at all.
> May be that is a bug but with wrong source input it not strange :)

This is _exactly_ what happens in the "make install" part, it builds to .info 
files write a third of zero length segfaults and then due to recursion i 
guess the parallel make segfaults to and the install process is aborted.

It succeeds on the gb.net.info and gb.qt.ext.info files, but segfualts on 
gb.qt.kde.html.info yielding a filesize of 292 bytes. But gb.sdl.info is 
written as well with zero-length thiugh.

Is the problem in the qt libraries? khtml, maybe? I hate to reconpile KDE, 
but, well it has been a while :-)

>
> I did a verify a few days back and I have a component build that is not
> ready and has somewhere a error so gbi abort fatal. removing this component
> and gbi goes 100% ok.
>
> Rob:
> There is nothing with building a rpm for distribution but one received from
> gentoo that maybe is mangled in bad way by them.
>
> Ron

Jesper