[RFA/RFC] Multilibs and gdb.asm

[RFA/RFC] Multilibs and gdb.asm

[Elena Zannoni]

In gdb.asm/asm-source.exp, the testsuite passes multilib options to
the assembler according to the compiler syntax.
Obviously this doesn't work, as already noted by Nick Clifton in:

http://sources.redhat.com/ml/gdb-patches/2002-01/msg00282.html

A solution wasn't reached at the time.

I found it useful to just bail out of the test if some multilibs were
detected. At least it reduced the noise in the testsuite results.

Is this too drastic?

Elena


2002-04-05  Elena Zannoni  <ezannoni@redhat.com>

	* gdb.asm/asm-source.exp: Bail out if multilibs are detected.


Index: asm-source.exp
===================================================================
RCS file: /cvs/uberbaum/gdb/testsuite/gdb.asm/asm-source.exp,v
retrieving revision 1.19
diff -u -p -r1.19 asm-source.exp
--- asm-source.exp	2002/04/05 02:48:23	1.19
+++ asm-source.exp	2002/04/05 21:15:04
@@ -67,10 +67,29 @@ if { "${asm-arch}" == "" } {
     gdb_suppress_entire_file "Assembly source test -- not implemented for this target."
 }
 
+# Watch out, we are invoking the assembler, but the testsuite sets multilib
+# switches according to compiler syntax.  If we pass these options straight
+# to the assembler, they won't always make sense.  If we don't pass them to
+# the assembler, the final link will complain that the object files were
+# built with different defaults.  So no matter what we do, we lose.  We may as
+# well get out of this test sooner rather than later.
+set dest [target_info name]
+if [board_info $dest exists multilib_flags] {
+	set multilib_flags [board_info $dest multilib_flags]
+	if { "${multilib_flags}" != "" } {
+	   gdb_suppress_entire_file "Assembly source test -- multilibs not supported by this test."
+	   return;
+	 }
+}
+
 set testfile "asm-source"

Re: [RFA/RFC] Multilibs and gdb.asm

[Daniel Jacobowitz]

On Fri, Apr 05, 2002 at 04:19:20PM -0500, Elena Zannoni wrote:
> 
> In gdb.asm/asm-source.exp, the testsuite passes multilib options to
> the assembler according to the compiler syntax.
> Obviously this doesn't work, as already noted by Nick Clifton in:
> 
> http://sources.redhat.com/ml/gdb-patches/2002-01/msg00282.html
> 
> A solution wasn't reached at the time.
> 
> I found it useful to just bail out of the test if some multilibs were
> detected. At least it reduced the noise in the testsuite results.
> 
> Is this too drastic?

I've got a silly suggestion.  Is there any reason not to assemble by
invoking the compiler, for this test?  Let it do all the multilib
footwork.

-- 
Daniel Jacobowitz                           Carnegie Mellon University
MontaVista Software                         Debian GNU/Linux Developer

Re: [RFA/RFC] Multilibs and gdb.asm

[Andrew Cagney]

> On Fri, Apr 05, 2002 at 04:19:20PM -0500, Elena Zannoni wrote:
> 
>> 
>> In gdb.asm/asm-source.exp, the testsuite passes multilib options to
>> the assembler according to the compiler syntax.
>> Obviously this doesn't work, as already noted by Nick Clifton in:
>> 
>> http://sources.redhat.com/ml/gdb-patches/2002-01/msg00282.html
>> 
>> A solution wasn't reached at the time.
>> 
>> I found it useful to just bail out of the test if some multilibs were
>> detected. At least it reduced the noise in the testsuite results.
>> 
>> Is this too drastic?
> 
> 
> I've got a silly suggestion.  Is there any reason not to assemble by
> invoking the compiler, for this test?  Let it do all the multilib
> footwork.

What compiler?  (There may be an assembler and linker but no compiler)

Andrew

Re: [RFA/RFC] Multilibs and gdb.asm

[Fernando Nasser]

This is very unfortunate, but if there is no way around it we
have no other choice.

Go ahead, please.

Fernando


Elena Zannoni wrote:
> 
> In gdb.asm/asm-source.exp, the testsuite passes multilib options to
> the assembler according to the compiler syntax.
> Obviously this doesn't work, as already noted by Nick Clifton in:
> 
> http://sources.redhat.com/ml/gdb-patches/2002-01/msg00282.html
> 
> A solution wasn't reached at the time.
> 
> I found it useful to just bail out of the test if some multilibs were
> detected. At least it reduced the noise in the testsuite results.
> 
> Is this too drastic?
> 
> Elena
> 
> 2002-04-05  Elena Zannoni  <ezannoni@redhat.com>
> 
>         * gdb.asm/asm-source.exp: Bail out if multilibs are detected.
> 
> Index: asm-source.exp
> ===================================================================
> RCS file: /cvs/uberbaum/gdb/testsuite/gdb.asm/asm-source.exp,v
> retrieving revision 1.19
> diff -u -p -r1.19 asm-source.exp
> --- asm-source.exp      2002/04/05 02:48:23     1.19
> +++ asm-source.exp      2002/04/05 21:15:04
> @@ -67,10 +67,29 @@ if { "${asm-arch}" == "" } {
>      gdb_suppress_entire_file "Assembly source test -- not implemented for this target."
>  }
> 
> +# Watch out, we are invoking the assembler, but the testsuite sets multilib
> +# switches according to compiler syntax.  If we pass these options straight
> +# to the assembler, they won't always make sense.  If we don't pass them to
> +# the assembler, the final link will complain that the object files were
> +# built with different defaults.  So no matter what we do, we lose.  We may as
> +# well get out of this test sooner rather than later.
> +set dest [target_info name]
> +if [board_info $dest exists multilib_flags] {
> +       set multilib_flags [board_info $dest multilib_flags]
> +       if { "${multilib_flags}" != "" } {
> +          gdb_suppress_entire_file "Assembly source test -- multilibs not supported by this test."
> +          return;
> +        }
> +}
> +
>  set testfile "asm-source"

-- 
Fernando Nasser
Red Hat Canada Ltd.                     E-Mail:  fnasser@redhat.com
2323 Yonge Street, Suite #300
Toronto, Ontario   M4P 2C9

Re: [RFA/RFC] Multilibs and gdb.asm

[Andrew Cagney]

> This is very unfortunate, but if there is no way around it we
> have no other choice.
> 
> Go ahead, please.
> 
> Fernando
> 

Hmm, suggest a bug report as well to track this.  Er, there is a bug 
report tracking it!  See gdb/285.

Add ``Work around PR gdb/285.'' to the changelog/commit message?

enjoy,
Andrew

Re: [RFA/RFC] Multilibs and gdb.asm

[Michael Snyder]

Andrew Cagney wrote:
> 
> > On Fri, Apr 05, 2002 at 04:19:20PM -0500, Elena Zannoni wrote:
> >
> >>
> >> In gdb.asm/asm-source.exp, the testsuite passes multilib options to
> >> the assembler according to the compiler syntax.
> >> Obviously this doesn't work, as already noted by Nick Clifton in:
> >>
> >> http://sources.redhat.com/ml/gdb-patches/2002-01/msg00282.html
> >>
> >> A solution wasn't reached at the time.
> >>
> >> I found it useful to just bail out of the test if some multilibs were
> >> detected. At least it reduced the noise in the testsuite results.
> >>
> >> Is this too drastic?
> >
> >
> > I've got a silly suggestion.  Is there any reason not to assemble by
> > invoking the compiler, for this test?  Let it do all the multilib
> > footwork.
> 
> What compiler?  (There may be an assembler and linker but no compiler)

No, Daniel's right.  At some point last year, we changed the test
so that it invokes "gdb_compile" to do the final link.  I think that
was because the compiler would figure out what libraries needed to
be linked in.  I've been vaguely worried about that change...

Re: [RFA/RFC] Multilibs and gdb.asm

[Michael Snyder]

Daniel Jacobowitz wrote:
> 
> On Fri, Apr 05, 2002 at 04:19:20PM -0500, Elena Zannoni wrote:
> >
> > In gdb.asm/asm-source.exp, the testsuite passes multilib options to
> > the assembler according to the compiler syntax.
> > Obviously this doesn't work, as already noted by Nick Clifton in:
> >
> > http://sources.redhat.com/ml/gdb-patches/2002-01/msg00282.html
> >
> > A solution wasn't reached at the time.
> >
> > I found it useful to just bail out of the test if some multilibs were
> > detected. At least it reduced the noise in the testsuite results.
> >
> > Is this too drastic?
> 
> I've got a silly suggestion.  Is there any reason not to assemble by
> invoking the compiler, for this test?  Let it do all the multilib
> footwork.

Oh, that would be a problem for targets where there is no compiler...

Re: [RFA/RFC] Multilibs and gdb.asm

[Elena Zannoni]

Andrew Cagney writes:
 > > This is very unfortunate, but if there is no way around it we
 > > have no other choice.
 > > 
 > > Go ahead, please.
 > > 
 > > Fernando
 > > 
 > 
 > Hmm, suggest a bug report as well to track this.  Er, there is a bug 
 > report tracking it!  See gdb/285.
 > 
 > Add ``Work around PR gdb/285.'' to the changelog/commit message?
 > 
 > enjoy,
 > Andrew
 > 

Committed.

Elena