Build with --disable-arch32-tests for gen-type-funit-tests

Build with --disable-arch32-tests for gen-type-funit-tests

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 210 bytes --]

Hi,

Currently the standard build on x86_64 is broken because
gen-type-funit-tests isn't properly generated in the arch32 dirs. For
now the workaround is to build with --disable-arch32.

Cheers,

Mark

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: Build with --disable-arch32-tests for gen-type-funit-tests

[Mark Wielaard]

[-- Attachment #1: Type: text/plain, Size: 951 bytes --]

On Wed, 2007-10-31 at 12:43 +0100, Mark Wielaard wrote:
> Hi,
> 
> Currently the standard build on x86_64 is broken because
> gen-type-funit-tests isn't properly generated in the arch32 dirs. For
> now the workaround is to build with --disable-arch32.

This should now all be fixed by:

commit b293ec8713bfc401241fc2dc0935fea152794438
Author: Andrew Cagney <cagney@redhat.com>
Date:   Wed Oct 31 17:20:55 2007 -0400

  For .c-in et.al., list base .c in SOURCES.
    
  frysk-common/ChangeLog
  2007-10-31  Andrew Cagney  <cagney@redhat.com>
    
  * Makefile.gen.sh (.c-in, .cxx-in, .c-sh, .cxx-sh): Add base .c or
  .cxx file to ...SOURCES and not the original source.  Drop use of
  automake's nodist_

http://sourceware.org/git/?p=frysk.git;a=commitdiff;h=b293ec8713bfc401241fc2dc0935fea152794438

Andrew, would it make sense to automatically run the arch32 tests in
make check whenever they are build?

Cheers,

Mark

[-- Attachment #2: This is a digitally signed message part --]
[-- Type: application/pgp-signature, Size: 189 bytes --]

Re: Build with --disable-arch32-tests for gen-type-funit-tests

[Andrew Cagney]

Mark Wielaard wrote:
> This should now all be fixed by:
>
(I'm assuming there wasn't a bug)
> Andrew, would it make sense to automatically run the arch32 tests in
> make check whenever they are build?
>   
Enabling 32-on-64 specific tests? Yes. They exercise 32-on-64 specific 
edge cases (the pedant will point out that when not run on a 64-bit 
system they should be marked as unresolved, not unsupported :-).  For 
instance testing 64-32-64 execs.  This is on my to-do list.

Running test-runner twice, once in 32-bit mode and once in 64-bit mode, 
is something better suited for the make file.

Andrew



> Cheers,
>
> Mark
>