Mail Archives: djgpp-workers/2001/08/24/07:29:57
On 24 Aug 2001, at 14:10, Eli Zaretskii wrote:
> > From: pavenis AT lanet DOT lv
> > Date: Fri, 24 Aug 2001 13:19:00 +0300
> > >
> > > I don't think so: that could mean a chicken-and-egg problem. For
> > > example, the script in the djlsr distribution might not work with the
> > > version of GCC used to build the library.
> >
> > Linker script is used by binutils not gcc.
>
> I know that. But it should be compatible with the compiler as well.
> IIRC, the reason GCC 3 has a linker script in its distribution is that
> some changes in the code it produces required changes in the linker
> script. So the script which comes with the compiler, if there is one,
> is guaranteed to work.
>
There were aditional sections and linker must know where to put them.
Earlier GCC versions didn't generate such sections, so no related
problems and also no breakage. Very old binutils is another thing
but they will not work with gcc-3.0.1 anyway.
> > > I think the best way is to use the script which comes with the
> > > compiler used to compile the library: it's guaranteed to work (if it
> > > doesn't, they don't have a working installation).
> >
> > There should be no problems if binutils is not too old
>
> We cannot guarantee anything about the CVS version: it's largely
> untested.
>
> Can you tell what's wrong with using the change I suggested?
If some package needs specific linker script, it is best that it provides
script (Currently out CVS version needs it when not very new binutils
versions are being used). Relaying on name which is now used by
GCC is not safe. It's seems unlikely I'll change name again, but anyway
I would like not to multiply references to djgpp-x.djl as it's only
temporary solution.
Andris
- Raw text -