More info on i386 alignment problem

[corsepiu at faw.uni-ulm.de (Ralf Corsepius)]

Stephan Wilms wrote:

> Hi again,
>
> I know I'm posting a lot today, but I might be catching on to a bug in
> the gcc "configure" skript here. The context is: I am trying to build
> EGCS for taget i386 on WinNT4 using CagWin 20.1. The "make cross" fails
> with the error message "Alignment not a power of 2".
>
> Let's see now: I am looking at the following file:
>      /source/egcs-1.1b/gcc/configure
>
> On line 5254 that script starts to check for assembler alignment
> capabilities. The problem seems to be that is does not seem to find an
> assembler. The first check is:
>      if [ -x as$host_exeext ]; then
> That is not successful because there is no assembler in the current
> directory (which at that point is "/build/egcs/gcc"). So the "if" block
> is ignored and the script continues with:
>      elif [ -f $gcc_cv_as_gas_srcdir/configure.in ]; then
> where "gcc_cv_as_gas_srcdir" holds the value "/source/egcs-1.1b/gas".
> This directory does not exist and therefore the test fails and the "elif"
> command block does not get executed. The last check is:
>      elif [ x$host = x$target ]; then
> which fails as well, because I am not doing a (real) native build
>
> Since all checks mentioned above fail (and I did not find any others) the
> script is now convinced that there is no assembler to be found and
> therefore does not test and set any assembler alignment options.
>

This analysis seems is correct, the work around is easy: Install
i386-rtems-gas and add the directory containing i386-rtems-gas[.exe] to
$PATH before configuring egcs.

Ralf


--
Ralf Corsepius
Forschungsinstitut fuer Anwendungsorientierte Wissensverarbeitung (FAW)
Helmholtzstr. 16, 89081 Ulm, Germany     Tel: +49/731/501-8690
mailto:corsepiu at faw.uni-ulm.de           FAX: +49/731/501-999
http://www.faw.uni-ulm.de