Alignment problem with i386 on WinNT

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

Stephan Wilms wrote:

> Hi,
>
> Remember my problems with the "Alignment not a power of 2" error message
> when trying to "makae cross" EGCS for WinNT ? Well, presently I am doing
> some research based on the configuration results for EGCS and I might be
> onto something here. Lets see ... What I do is this (keep in mind: it's
> CygWin 20.1 on WinNT 4):
>

:=)

>
>    cd /build/egcs
>    /source/configure --verbose --target=i386-rtems  \
>                      --prefix=/gcc-i386-rtems       \
>                      --with-gnu-as --with-gnu-ld    \
>                      --with-newlib
>
> The result is hundreds of lines of output, most of them beginning with
> "checking", but I noted something special. There comes along a line that
> says "Configuring gcc...", lots of "Checking" below that, and the very
> last line of checking says: "checking assembler alignment features...".
>

Check the config.log files in your build-tree. This should provide further
information about what's going on here.

>
> What puzzles me is that there is nothing listed after the three dots.
> Does this mean that the configure script was not able to determine the
> alignment features for the "i386" target ?

Possible, but unlikely. Normally this simply indicates that the check was
not completed or the configure scripts output was corrupted (e.g. i/o
redirection became corrupted).

> Right now I am looking at the
> file "source/egcs-1.1b/configure" where I found the alignment checking on
> line (5251 and following). It looks like there is an "echo" of results if
> alignment modes were determined.
>
> In my case the resultsof the alignment checking must be empty because all
> I get as output is this line that I already quoted above:
>    "checking assembler alignment features...
>
> Now I am going to have a look at the "confdefs.h" and "" files to see iof
> they offer any further insight. I'll keep you updated if I find anything.
>

If I understand correctly, egcs/gcc/configure does not terminate at this
point. This normally indicates that something inside the configure shell
script went wrong very badly, which isn't necessarily related to the purpose
of the check actually being executed in configure.

In most cases this normally happens when an auxillary program used inside
configure scripts (e.g. sed, test), wait for input on stdin but do not
receive it.

On Cygwin I wouldn't be astonished to see this caused by \\ in environment
variables, by "|" (pipe) dumping core (AFAIK, cygwin has problems with
pipes), or by having added "\\" to configure scripts using DOS editors or by
mounting partitions in text mode introducing CRLFs to configure scripts.

>
> Can anyone offer some advice or information on the subject of alignment
> for the "i386" target ?

Though I can't exclude that the check for alignment in gcc/configure.in is
broken, I tend to think that it is unlikely to be the cause of your
problems.

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