public inbox for binutils@sourceware.org
 help / color / mirror / Atom feed
* Re: [Patch] sh64: Fix gas testsuite expected output
@ 2002-10-15 20:16 Stephen Clarke
  2002-10-15 20:27 ` H. J. Lu
  2002-10-15 21:51 ` [Patch] sh64: Fix gas testsuite expected output Hans-Peter Nilsson
  0 siblings, 2 replies; 42+ messages in thread
From: Stephen Clarke @ 2002-10-15 20:16 UTC (permalink / raw)
  To: H. J. Lu; +Cc: binutils

From: "Stephen Clarke"
Sent: Monday, October 14, 2002 10:05 PM
>
> From: "H. J. Lu"
> Sent: Monday, October 14, 2002 9:38 PM
> 
> > [long list of further failures]
> 
> I don't get any of these.  I'll try to work out why.
> It could help if you'd send me the gas.log file.

It looks like these failures are because of:

.../objdump: dump.o: File format is ambiguous
.../objdump: Matching formats: elf32-sh64big-linux elf32-sh64

and that's because both sh-linux and sh64-linux targets
are enabled, i.e. binutils is configured with
--enable-targets=...,sh-linux,sh64-linux,...

The definition of sh-*-linux* in config.bfd is:

sh-*-linux*)
    targ_defvec=bfd_elf32_shblin_vec
    targ_selvecs=bfd_elf32_shlin_vec
#ifdef BFD64
    targ_selvecs="${targ_selvecs} bfd_elf32_sh64_vec bfd_elf32_sh64l_vec bfd_elf64_sh64_vec bfd_elf64_sh64l_vec"
#endif

This includes some "sh64" target vectors, which then cannot be distinguished
from sh64-linux target vectors.

I'm not sure why sh-*-linux* should include sh64 target vectors.  My understanding
of "sh" is that it is supposed to be 32-bit sh targets only, and it
should avoid any overhead required for 64-bit targets.

I suggest that the sh64 vectors are removed from sh-*-linux*.
What do the maintainers think?

(Aside: as mentioned in http://sources.redhat.com/ml/binutils/2002-08/msg00069.html
this particular use of ifdef BFD64 has no effect in config.bfd: the sh64 vectors are
built into an sh-linux target whether or not --enable-64-bit-bfd is specified.  I intended to
submit a patch to strip out the ineffectual ifdefs, but forgot.)

Steve.


__________________________________________________
Do You Yahoo!?
Sign up for SBC Yahoo! Dial - First Month Free
http://sbc.yahoo.com

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-15 20:16 [Patch] sh64: Fix gas testsuite expected output Stephen Clarke
@ 2002-10-15 20:27 ` H. J. Lu
  2002-10-15 21:39   ` Stephen Clarke
  2002-10-15 21:51 ` [Patch] sh64: Fix gas testsuite expected output Hans-Peter Nilsson
  1 sibling, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-15 20:27 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Tue, Oct 15, 2002 at 03:47:44PM -0700, Stephen Clarke wrote:
> From: "Stephen Clarke"
> Sent: Monday, October 14, 2002 10:05 PM
> >
> > From: "H. J. Lu"
> > Sent: Monday, October 14, 2002 9:38 PM
> > 
> > > [long list of further failures]
> > 
> > I don't get any of these.  I'll try to work out why.
> > It could help if you'd send me the gas.log file.
> 
> It looks like these failures are because of:
> 
> .../objdump: dump.o: File format is ambiguous
> .../objdump: Matching formats: elf32-sh64big-linux elf32-sh64
> 
> and that's because both sh-linux and sh64-linux targets
> are enabled, i.e. binutils is configured with
> --enable-targets=...,sh-linux,sh64-linux,...
> 

That is not right. Something is wrong with ELF/SH. Can you tell me
the difference between elf32-sh64big-linux and elf32-sh64 in the
ELF header?



H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-15 20:27 ` H. J. Lu
@ 2002-10-15 21:39   ` Stephen Clarke
  2002-10-15 22:32     ` H. J. Lu
  0 siblings, 1 reply; 42+ messages in thread
From: Stephen Clarke @ 2002-10-15 21:39 UTC (permalink / raw)
  To: H. J. Lu; +Cc: binutils

From: "H. J. Lu"
Sent: Tuesday, October 15, 2002 8:27 PM
> 
> That is not right. Something is wrong with ELF/SH. Can you tell me
> the difference between elf32-sh64big-linux and elf32-sh64 in the
> ELF header?

I can't: I don't think there is a difference in the ELF header.
The same issue applies to non-sh64 targets,
e.g.elf32-shl and elf32-sh-linux.


__________________________________________________
Do You Yahoo!?
Sign up for SBC Yahoo! Dial - First Month Free
http://sbc.yahoo.com

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-15 20:16 [Patch] sh64: Fix gas testsuite expected output Stephen Clarke
  2002-10-15 20:27 ` H. J. Lu
@ 2002-10-15 21:51 ` Hans-Peter Nilsson
  1 sibling, 0 replies; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-15 21:51 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: H. J. Lu, binutils

On Tue, 15 Oct 2002, Stephen Clarke wrote:
> I'm not sure why sh-*-linux* should include sh64 target vectors.  My understanding
> of "sh" is that it is supposed to be 32-bit sh targets only, and it
> should avoid any overhead required for 64-bit targets.

For "sh" as opposed to "sh4" an other variants, the default is
to include the sh64 vectors, so sh-*-* tools can identify and
interpret sh64 objects.  GDB folks in particular like this.

> I suggest that the sh64 vectors are removed from sh-*-linux*.
> What do the maintainers think?

No, we've been through that before, summarized above.

As H.J. noted, the ambiguity bug is not solved by excluding
vectors.  The elf32-sh64big-linux and elf32-sh64 targets must be
disambiguated somehow; there should be some ELF header info that
tells them apart.  If there is nothing to tell them apart,
something should be added.

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-15 21:39   ` Stephen Clarke
@ 2002-10-15 22:32     ` H. J. Lu
  2002-10-15 22:54       ` Stephen Clarke
  0 siblings, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-15 22:32 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Tue, Oct 15, 2002 at 09:39:29PM -0700, Stephen Clarke wrote:
> From: "H. J. Lu"
> Sent: Tuesday, October 15, 2002 8:27 PM
> > 
> > That is not right. Something is wrong with ELF/SH. Can you tell me
> > the difference between elf32-sh64big-linux and elf32-sh64 in the
> > ELF header?
> 
> I can't: I don't think there is a difference in the ELF header.
> The same issue applies to non-sh64 targets,
> e.g.elf32-shl and elf32-sh-linux.
> 

Then, I have to ask, what are the differences between elf32-shl and
elf32-sh-linux? If you tell me that, I may come up with something.


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-15 22:32     ` H. J. Lu
@ 2002-10-15 22:54       ` Stephen Clarke
  2002-10-15 23:00         ` H. J. Lu
  0 siblings, 1 reply; 42+ messages in thread
From: Stephen Clarke @ 2002-10-15 22:54 UTC (permalink / raw)
  To: H. J. Lu; +Cc: binutils

From: "H. J. Lu" <hjl@lucon.org>
Sent: Tuesday, October 15, 2002 10:32 PM


> On Tue, Oct 15, 2002 at 09:39:29PM -0700, Stephen Clarke wrote:
> > From: "H. J. Lu"
> > Sent: Tuesday, October 15, 2002 8:27 PM
> > > 
> > > That is not right. Something is wrong with ELF/SH. Can you tell me
> > > the difference between elf32-sh64big-linux and elf32-sh64 in the
> > > ELF header?
> > 
> > I can't: I don't think there is a difference in the ELF header.
> > The same issue applies to non-sh64 targets,
> > e.g.elf32-shl and elf32-sh-linux.
> > 
> 
> Then, I have to ask, what are the differences between elf32-shl and
> elf32-sh-linux? If you tell me that, I may come up with something.

If you mean what are the differences in the ELF header between
elf32-shl and elf32-sh-linux, then I don't think there are any.   If I
build a binutils with these two targets enabled, then objdump gives
error messages about ambiguous file format:
sh-elf-objdump: test.o: File format is ambiguous
sh-elf-objdump: Matching formats: elf32-shl elf32-sh-linux



__________________________________________________
Do you Yahoo!?
Faith Hill - Exclusive Performances, Videos & More
http://faith.yahoo.com

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-15 22:54       ` Stephen Clarke
@ 2002-10-15 23:00         ` H. J. Lu
  2002-10-15 23:39           ` H. J. Lu
  0 siblings, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-15 23:00 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Tue, Oct 15, 2002 at 10:51:59PM -0700, Stephen Clarke wrote:
> From: "H. J. Lu" <hjl@lucon.org>
> Sent: Tuesday, October 15, 2002 10:32 PM
> 
> 
> > On Tue, Oct 15, 2002 at 09:39:29PM -0700, Stephen Clarke wrote:
> > > From: "H. J. Lu"
> > > Sent: Tuesday, October 15, 2002 8:27 PM
> > > > 
> > > > That is not right. Something is wrong with ELF/SH. Can you tell me
> > > > the difference between elf32-sh64big-linux and elf32-sh64 in the
> > > > ELF header?
> > > 
> > > I can't: I don't think there is a difference in the ELF header.
> > > The same issue applies to non-sh64 targets,
> > > e.g.elf32-shl and elf32-sh-linux.
> > > 
> > 
> > Then, I have to ask, what are the differences between elf32-shl and
> > elf32-sh-linux? If you tell me that, I may come up with something.
> 
> If you mean what are the differences in the ELF header between
> elf32-shl and elf32-sh-linux, then I don't think there are any.   If I

No, why do you have elf32-shl and elf32-sh-linux? Can you just use
elf32-shl for Linux?


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-15 23:00         ` H. J. Lu
@ 2002-10-15 23:39           ` H. J. Lu
  2002-10-16  1:47             ` Alan Modra
  0 siblings, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-15 23:39 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Tue, Oct 15, 2002 at 11:00:20PM -0700, H. J. Lu wrote:
> On Tue, Oct 15, 2002 at 10:51:59PM -0700, Stephen Clarke wrote:
> > From: "H. J. Lu" <hjl@lucon.org>
> > Sent: Tuesday, October 15, 2002 10:32 PM
> > 
> > 
> > > On Tue, Oct 15, 2002 at 09:39:29PM -0700, Stephen Clarke wrote:
> > > > From: "H. J. Lu"
> > > > Sent: Tuesday, October 15, 2002 8:27 PM
> > > > > 
> > > > > That is not right. Something is wrong with ELF/SH. Can you tell me
> > > > > the difference between elf32-sh64big-linux and elf32-sh64 in the
> > > > > ELF header?
> > > > 
> > > > I can't: I don't think there is a difference in the ELF header.
> > > > The same issue applies to non-sh64 targets,
> > > > e.g.elf32-shl and elf32-sh-linux.
> > > > 
> > > 
> > > Then, I have to ask, what are the differences between elf32-shl and
> > > elf32-sh-linux? If you tell me that, I may come up with something.
> > 
> > If you mean what are the differences in the ELF header between
> > elf32-shl and elf32-sh-linux, then I don't think there are any.   If I
> 
> No, why do you have elf32-shl and elf32-sh-linux? Can you just use
> elf32-shl for Linux?
> 

It looks like Linux/sh uses a different ABI than ELF/sh. Please
consider setting ELFOSABI_LINUX. See elf32-hppa.c/elf64-hppa.c
for how to do it. You should do it ASAP before it is too late.


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-15 23:39           ` H. J. Lu
@ 2002-10-16  1:47             ` Alan Modra
  2002-10-16  6:16               ` H. J. Lu
  0 siblings, 1 reply; 42+ messages in thread
From: Alan Modra @ 2002-10-16  1:47 UTC (permalink / raw)
  To: H. J. Lu; +Cc: Stephen Clarke, binutils

On Tue, Oct 15, 2002 at 11:39:36PM -0700, H. J. Lu wrote:
> It looks like Linux/sh uses a different ABI than ELF/sh. Please
> consider setting ELFOSABI_LINUX.

Hmm, yeah.  For SH, you have the added complication of NetBSD and QNX
targets.  The NetBSD one is virtually identical to the Linux one.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-16  1:47             ` Alan Modra
@ 2002-10-16  6:16               ` H. J. Lu
  2002-10-16  8:03                 ` Alan Modra
  0 siblings, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-16  6:16 UTC (permalink / raw)
  To: Stephen Clarke, binutils

On Wed, Oct 16, 2002 at 06:17:22PM +0930, Alan Modra wrote:
> On Tue, Oct 15, 2002 at 11:39:36PM -0700, H. J. Lu wrote:
> > It looks like Linux/sh uses a different ABI than ELF/sh. Please
> > consider setting ELFOSABI_LINUX.
> 
> Hmm, yeah.  For SH, you have the added complication of NetBSD and QNX
> targets.  The NetBSD one is virtually identical to the Linux one.
> 

That reminds me that QNX needs to get their own ELFOSABI_QNX since
they do have a different ABI, even on x86. If NetBSD uses the same
ABI as Linux, they should share the same ELFOSABI_XXX.


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-16  6:16               ` H. J. Lu
@ 2002-10-16  8:03                 ` Alan Modra
  2002-10-16  8:14                   ` H. J. Lu
  2002-10-16  9:50                   ` Matt Thomas
  0 siblings, 2 replies; 42+ messages in thread
From: Alan Modra @ 2002-10-16  8:03 UTC (permalink / raw)
  To: H. J. Lu; +Cc: Stephen Clarke, binutils

On Wed, Oct 16, 2002 at 06:16:25AM -0700, H. J. Lu wrote:
> If NetBSD uses the same
> ABI as Linux, they should share the same ELFOSABI_XXX.

Aye, but should NetBSD use ELFOSABI_LINUX or Linux use ELFOSABI_NETBSD?
Bun fight coming up..

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-16  8:03                 ` Alan Modra
@ 2002-10-16  8:14                   ` H. J. Lu
  2002-10-16  9:50                   ` Matt Thomas
  1 sibling, 0 replies; 42+ messages in thread
From: H. J. Lu @ 2002-10-16  8:14 UTC (permalink / raw)
  To: Stephen Clarke, binutils

On Thu, Oct 17, 2002 at 12:33:04AM +0930, Alan Modra wrote:
> On Wed, Oct 16, 2002 at 06:16:25AM -0700, H. J. Lu wrote:
> > If NetBSD uses the same
> > ABI as Linux, they should share the same ELFOSABI_XXX.
> 
> Aye, but should NetBSD use ELFOSABI_LINUX or Linux use ELFOSABI_NETBSD?
> Bun fight coming up..
> 

I took a look. I believe the straight ELF/sh should use

#define ELFOSABI_STANDALONE 255 /* Standalone (embedded) application */

and we delete Linux/sh and NetBSD/sh. Then we create Unix/sh with

#define ELFOSABI_NONE   0       /* UNIX System V ABI */

I hope it is not late to do the right thing. That means there is no
change in Linux/sh/NetBSD/sh binaries. But ELF/sh will be different.


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-16  8:03                 ` Alan Modra
  2002-10-16  8:14                   ` H. J. Lu
@ 2002-10-16  9:50                   ` Matt Thomas
  2002-10-18  1:56                     ` Alan Modra
  1 sibling, 1 reply; 42+ messages in thread
From: Matt Thomas @ 2002-10-16  9:50 UTC (permalink / raw)
  To: Alan Modra; +Cc: binutils

At 08:03 AM 10/16/2002, Alan Modra wrote:
>On Wed, Oct 16, 2002 at 06:16:25AM -0700, H. J. Lu wrote:
> > If NetBSD uses the same
> > ABI as Linux, they should share the same ELFOSABI_XXX.
>
>Aye, but should NetBSD use ELFOSABI_LINUX or Linux use ELFOSABI_NETBSD?
>Bun fight coming up..

IMO, even though Linux and NetBSD use the same calling conventions,
they are different ABIs since syscalls are handled differently and
libc has different conventions.  You can't use a Linux shared lib
natively on NetBSD nor can you use a NetBSD shared lib natively on
Linux.  Without a kernel compatibility layer, a small staticly
linked "Hello World" program linked on one O/S won't run on the
other.




-- 
Matt Thomas               Internet:   matt@3am-software.com
3am Software Foundry      WWW URL:    http://www.3am-software.com/bio/matt/
Cupertino, CA             Disclaimer: I avow all knowledge of this message

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-16  9:50                   ` Matt Thomas
@ 2002-10-18  1:56                     ` Alan Modra
  2002-10-18 10:22                       ` Ulrich Drepper
  0 siblings, 1 reply; 42+ messages in thread
From: Alan Modra @ 2002-10-18  1:56 UTC (permalink / raw)
  To: Matt Thomas; +Cc: binutils

On Wed, Oct 16, 2002 at 09:48:15AM -0700, Matt Thomas wrote:
> IMO, even though Linux and NetBSD use the same calling conventions,
> they are different ABIs since syscalls are handled differently and
> libc has different conventions.  You can't use a Linux shared lib
> natively on NetBSD nor can you use a NetBSD shared lib natively on
> Linux.  Without a kernel compatibility layer, a small staticly
> linked "Hello World" program linked on one O/S won't run on the
> other.

This argument is irrelevant to correct usage of EI_OSABI.  For
example, it's quite easy to create an app on a recent version of Linux
that won't run on an older Linux.  Does that mean OSABI or ABIVERSION
needs to change?  Not at all.

I believe the real test for a change in OSABI or ABIVERSION is:
Will a consumer of ELF files, such as a linker, that properly handles
ELF files conforming to the gABI and relevant psABI, be able to
properly handle your particular ELF files?  ie.  OSABI and ABIVERSION
are used in cases where there are variations and/or additions to
fields specified by the ELF spec, and has nothing to do with say, the
contents of the .text section.  Typically, you'll need to change OSABI
if you use any of the various *_LOOS to *_HIOS definitions, or use new
or incompatible relocation types.  Search the binutils list for lots
of discussion on this.  ;-)  One particularly relevant post is
http://sources.redhat.com/ml/binutils/2000-11/msg00383.html

That said, I'm not religiously opposed to the BSD's making
non-standard use of OSABI.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-18  1:56                     ` Alan Modra
@ 2002-10-18 10:22                       ` Ulrich Drepper
  2002-10-18 10:38                         ` H. J. Lu
  2002-10-18 12:33                         ` Hans-Peter Nilsson
  0 siblings, 2 replies; 42+ messages in thread
From: Ulrich Drepper @ 2002-10-18 10:22 UTC (permalink / raw)
  To: Alan Modra; +Cc: Matt Thomas, binutils

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Alan Modra wrote:

> I believe the real test for a change in OSABI or ABIVERSION is:
> Will a consumer of ELF files, such as a linker, that properly handles
> ELF files conforming to the gABI and relevant psABI, be able to
> properly handle your particular ELF files?


As said many times, yes, this is the only use for this field.  All other
marking of an ELF file must be done with other means and the method
agreed on a looong time ago is a special note segment.

- -- 
- --------------.                        ,-.            444 Castro Street
Ulrich Drepper \    ,-----------------'   \ Mountain View, CA 94041 USA
Red Hat         `--' drepper at redhat.com `---------------------------
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.0.7 (GNU/Linux)

iD8DBQE9sENb2ijCOnn/RHQRAq76AKDDHCcZHA1oIhSuteLT4ddj5yW/4QCghoy+
6nCoSjWSmg1CEPg5QWm85Q4=
=x2qI
-----END PGP SIGNATURE-----

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-18 10:22                       ` Ulrich Drepper
@ 2002-10-18 10:38                         ` H. J. Lu
  2002-10-18 11:21                           ` Jason R Thorpe
  2002-10-18 12:33                         ` Hans-Peter Nilsson
  1 sibling, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-18 10:38 UTC (permalink / raw)
  To: Ulrich Drepper; +Cc: Alan Modra, Matt Thomas, binutils

On Fri, Oct 18, 2002 at 10:22:35AM -0700, Ulrich Drepper wrote:
> -----BEGIN PGP SIGNED MESSAGE-----
> Hash: SHA1
> 
> Alan Modra wrote:
> 
> > I believe the real test for a change in OSABI or ABIVERSION is:
> > Will a consumer of ELF files, such as a linker, that properly handles
> > ELF files conforming to the gABI and relevant psABI, be able to
> > properly handle your particular ELF files?
> 
> 
> As said many times, yes, this is the only use for this field.  All other
> marking of an ELF file must be done with other means and the method
> agreed on a looong time ago is a special note segment.
> 
 
As far as I know, ELF is an ABI for SVR4. That basically means it is
for Unix alike systems similar to SVR4. In elf32-sh.c, there are

#define ELF_MAXPAGESIZE         128

#define elf_symbol_leading_char '_'

They don't like a reasonable SVR4 system to me. It is more like a
standalone or embedded ABI. That is why I suggested ELF/sh should set

#define ELFOSABI_STANDALONE 255 /* Standalone (embedded) application */

We should create elf*-sh*-svr.c for SVR4 and use

#define ELFOSABI_NONE   0       /* UNIX System V ABI */

We should do it now before it is too late.


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-18 10:38                         ` H. J. Lu
@ 2002-10-18 11:21                           ` Jason R Thorpe
  2002-10-18 11:27                             ` Jason R Thorpe
  0 siblings, 1 reply; 42+ messages in thread
From: Jason R Thorpe @ 2002-10-18 11:21 UTC (permalink / raw)
  To: H. J. Lu; +Cc: Ulrich Drepper, Alan Modra, Matt Thomas, binutils

On Fri, Oct 18, 2002 at 10:38:11AM -0700, H. J. Lu wrote:

 > They don't like a reasonable SVR4 system to me. It is more like a
 > standalone or embedded ABI. That is why I suggested ELF/sh should set
 > 
 > #define ELFOSABI_STANDALONE 255 /* Standalone (embedded) application */
 > 
 > We should create elf*-sh*-svr.c for SVR4 and use
 > 
 > #define ELFOSABI_NONE   0       /* UNIX System V ABI */
 > 
 > We should do it now before it is too late.

HJ's suggestion seems reasonable to me.  Certainly the '-' prefix on the
symbol names in the "stock" SH ELF make the objects incompatible with
ELF objects from Unix systems.

(Though, instead of "-svr", I'd suggest "-unix" or "-unx" as the BFD
target suffix :-)

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-18 11:21                           ` Jason R Thorpe
@ 2002-10-18 11:27                             ` Jason R Thorpe
  0 siblings, 0 replies; 42+ messages in thread
From: Jason R Thorpe @ 2002-10-18 11:27 UTC (permalink / raw)
  To: H. J. Lu, Ulrich Drepper, Alan Modra, Matt Thomas, binutils

On Fri, Oct 18, 2002 at 11:21:16AM -0700, Jason R Thorpe wrote:

 > HJ's suggestion seems reasonable to me.  Certainly the '-' prefix on the
							  ^^^
						    oops, '_', of course

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-18 10:22                       ` Ulrich Drepper
  2002-10-18 10:38                         ` H. J. Lu
@ 2002-10-18 12:33                         ` Hans-Peter Nilsson
  2002-10-18 18:04                           ` Alan Modra
  2002-10-22  1:44                           ` kaz Kojima
  1 sibling, 2 replies; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-18 12:33 UTC (permalink / raw)
  To: binutils; +Cc: Alan Modra, Matt Thomas, Ulrich Drepper

On Fri, 18 Oct 2002, Ulrich Drepper wrote:
> Alan Modra wrote:
> > I believe the real test for a change in OSABI or ABIVERSION is:
> > Will a consumer of ELF files, such as a linker, that properly handles
> > ELF files conforming to the gABI and relevant psABI, be able to
> > properly handle your particular ELF files?
> As said many times, yes, this is the only use for this field.  All other
> marking of an ELF file must be done with other means and the method
> agreed on a looong time ago is a special note segment.

Sure, a .note section with identifying contents for a linked
object, but that doesn't solve it for a REL (meaning non-linked)
object.

Or perhaps you didn't really mean the .note section?

Suggestion: for REL (still meaning unlinked) objects have an
empty section with a special name (keyed on machine-name and
vendor name, like GNU as in GNU binutils, and on function) in
each unlinked object, e.g.
 .SH.GNU.abi.gnulinux
 .SH.GNU.abi.standalone
 .SH.GNU.abi.netbsd
(Names TBD; just examples.)  Caveat: it doesn't solve the
ambiguity for existing objects.  I think we can live with that,
if objects without such a section are treated as compatible with
whatever context they are used in.  This should keep backward
compatibility, and above all, old tools won't hork when they see
"new" objects.

Thoughts?

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-18 12:33                         ` Hans-Peter Nilsson
@ 2002-10-18 18:04                           ` Alan Modra
  2002-10-18 18:46                             ` Hans-Peter Nilsson
  2002-10-22  1:44                           ` kaz Kojima
  1 sibling, 1 reply; 42+ messages in thread
From: Alan Modra @ 2002-10-18 18:04 UTC (permalink / raw)
  To: Hans-Peter Nilsson; +Cc: binutils, Matt Thomas, Ulrich Drepper

On Fri, Oct 18, 2002 at 03:33:10PM -0400, Hans-Peter Nilsson wrote:
> Sure, a .note section with identifying contents for a linked
> object, but that doesn't solve it for a REL (meaning non-linked)
> object.
> 
> Or perhaps you didn't really mean the .note section?

.note works fine for object files too.  In fact, we already use them.
See gas/config/obj-elf.c:obj_elf_version and sco_id.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-18 18:04                           ` Alan Modra
@ 2002-10-18 18:46                             ` Hans-Peter Nilsson
  0 siblings, 0 replies; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-18 18:46 UTC (permalink / raw)
  To: Alan Modra; +Cc: binutils, Matt Thomas, Ulrich Drepper

On Sat, 19 Oct 2002, Alan Modra wrote:
> On Fri, Oct 18, 2002 at 03:33:10PM -0400, Hans-Peter Nilsson wrote:
> > Sure, a .note section with identifying contents for a linked
> > object, but that doesn't solve it for a REL (meaning non-linked)
> > object.
> >
> > Or perhaps you didn't really mean the .note section?
>
> .note works fine for object files too.  In fact, we already use them.
> See gas/config/obj-elf.c:obj_elf_version and sco_id.

Right.  Unfortunately they take up room (in the binary) and
accumulate in size when linking.  (SEC_MERGE on that section,
anyone?)  For *this* case, I'd prefer an empty section with a
magic name.

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-18 12:33                         ` Hans-Peter Nilsson
  2002-10-18 18:04                           ` Alan Modra
@ 2002-10-22  1:44                           ` kaz Kojima
  2002-10-22 10:14                             ` Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output) H. J. Lu
  1 sibling, 1 reply; 42+ messages in thread
From: kaz Kojima @ 2002-10-22  1:44 UTC (permalink / raw)
  To: hp; +Cc: binutils

Hans-Peter Nilsson <hp@bitrange.com> wrote:
> Suggestion: for REL (still meaning unlinked) objects have an
> empty section with a special name (keyed on machine-name and
> vendor name, like GNU as in GNU binutils, and on function) in
> each unlinked object, e.g.
>  .SH.GNU.abi.gnulinux
>  .SH.GNU.abi.standalone
>  .SH.GNU.abi.netbsd
> (Names TBD; just examples.)  Caveat: it doesn't solve the
> ambiguity for existing objects.  I think we can live with that,
> if objects without such a section are treated as compatible with
> whatever context they are used in.  This should keep backward
> compatibility, and above all, old tools won't hork when they see
> "new" objects.

I looked the original problem again. I got "File format is ambiguous"
messages when running gas testsuite for the sh4-unknown-linux-gnu
target configured with --enable-targets=all, but the error occurs
only for the big endian objects.
I found that bfd/format.c:bfd_check_format_matches doesn't fail when
the default vector is an element of the set of matching vectors. And
all FAILs are gone when changing bfd_check_format_matches so as to
permit also the alternative vector of the default one.

For sh64-unknown-linux-gnu, I got 37 unexpected failures after this
change. In these cases, bfd_check_format_matches failed for 64-bit
objects because the default format of sh64-unknown-linux-gnu target
is 32-bit elf32-sh64-linux format but matching formats are 64-bit
formats.

So I guess that the ambiguity for existing objects can be handled by
adding something to bfd_check_format_matches which selects an appropriate
representative vector from the ambiguity set if possible, as it was
already done when the default vector is included in this set.
Thoughts?

Regards,
	kaz

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output)
  2002-10-22  1:44                           ` kaz Kojima
@ 2002-10-22 10:14                             ` H. J. Lu
  2002-10-22 15:58                               ` kaz Kojima
  0 siblings, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-22 10:14 UTC (permalink / raw)
  To: kaz Kojima; +Cc: hp, binutils

On Tue, Oct 22, 2002 at 05:21:38PM +0900, kaz Kojima wrote:
> 
> For sh64-unknown-linux-gnu, I got 37 unexpected failures after this
> change. In these cases, bfd_check_format_matches failed for 64-bit
> objects because the default format of sh64-unknown-linux-gnu target
> is 32-bit elf32-sh64-linux format but matching formats are 64-bit
> formats.
> 
> So I guess that the ambiguity for existing objects can be handled by
> adding something to bfd_check_format_matches which selects an appropriate
> representative vector from the ambiguity set if possible, as it was
> already done when the default vector is included in this set.
> Thoughts?

Someone from SH should read

http://sources.redhat.com/ml/binutils/2002-10/msg00437.html

The old one is very unreasonable. Do it now before it is too late. I
can supply patches if necessary.


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output)
  2002-10-22 10:14                             ` Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output) H. J. Lu
@ 2002-10-22 15:58                               ` kaz Kojima
  2002-10-22 16:43                                 ` Hans-Peter Nilsson
  0 siblings, 1 reply; 42+ messages in thread
From: kaz Kojima @ 2002-10-22 15:58 UTC (permalink / raw)
  To: hjl; +Cc: hp, binutils

"H. J. Lu" <hjl@lucon.org> wrote:
>> So I guess that the ambiguity for existing objects can be handled by
>> adding something to bfd_check_format_matches which selects an appropriate
>> representative vector from the ambiguity set if possible, as it was
>> already done when the default vector is included in this set.
>> Thoughts?
> 
> Someone from SH should read
> 
> http://sources.redhat.com/ml/binutils/2002-10/msg00437.html
> 
> The old one is very unreasonable. Do it now before it is too late. I
> can supply patches if necessary.

I'm not a one from SH :-)
Anyway, my argument about bfd_check_format_matches is for
"existing objects".

Regards,
	kaz

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output)
  2002-10-22 15:58                               ` kaz Kojima
@ 2002-10-22 16:43                                 ` Hans-Peter Nilsson
  2002-10-22 22:55                                   ` Jason R Thorpe
  2002-10-23  5:12                                   ` kaz Kojima
  0 siblings, 2 replies; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-22 16:43 UTC (permalink / raw)
  To: kaz Kojima; +Cc: hjl, binutils

On Wed, 23 Oct 2002, kaz Kojima wrote:
> "H. J. Lu" <hjl@lucon.org> wrote:
> >> So I guess that the ambiguity for existing objects can be handled by
> >> adding something to bfd_check_format_matches which selects an appropriate
> >> representative vector from the ambiguity set if possible, as it was
> >> already done when the default vector is included in this set.
> >> Thoughts?

From my now-outside perspective, I think it'd be better to have
the port see a difference (which there currently is none for
*unlinked objects*), such as with my special-empty-section
proposal.  (It'd also be much less of a hassle than IMHO
wrongfully changing ELFOSABI.)

> > Someone from SH should read
> >
> > http://sources.redhat.com/ml/binutils/2002-10/msg00437.html
> >
> > The old one is very unreasonable. Do it now before it is too late. I
> > can supply patches if necessary.

FWIW, I disagree with H.J. about the gravity and the solution.

> I'm not a one from SH :-)

Me neither. :-)  I think Joern is the right person to ask for
that.

> Anyway, my argument about bfd_check_format_matches is for
> "existing objects".

Um, only for objects that aren't matched by the default target
vector, right?

For executables, make elf_object_p look at .note.ABI-tag. (They
don't have the same contents on NetBSD, no?)  Create one (or
something equivalent) for DSO:s too; it seems like that doesn't
happen.

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output)
  2002-10-22 16:43                                 ` Hans-Peter Nilsson
@ 2002-10-22 22:55                                   ` Jason R Thorpe
  2002-10-22 22:58                                     ` H. J. Lu
  2002-10-23  5:12                                   ` kaz Kojima
  1 sibling, 1 reply; 42+ messages in thread
From: Jason R Thorpe @ 2002-10-22 22:55 UTC (permalink / raw)
  To: Hans-Peter Nilsson; +Cc: kaz Kojima, hjl, binutils

On Tue, Oct 22, 2002 at 07:43:54PM -0400, Hans-Peter Nilsson wrote:

 > For executables, make elf_object_p look at .note.ABI-tag. (They
 > don't have the same contents on NetBSD, no?)  Create one (or
 > something equivalent) for DSO:s too; it seems like that doesn't
 > happen.

NetBSD uses .note.netbsd.ident ... and it's provided by the NetBSD
crtstuff, and so only applies to linked objects (programs, shared
libraries, etc.)

-- 
        -- Jason R. Thorpe <thorpej@wasabisystems.com>

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output)
  2002-10-22 22:55                                   ` Jason R Thorpe
@ 2002-10-22 22:58                                     ` H. J. Lu
  0 siblings, 0 replies; 42+ messages in thread
From: H. J. Lu @ 2002-10-22 22:58 UTC (permalink / raw)
  To: Jason R Thorpe, Hans-Peter Nilsson, kaz Kojima, binutils

On Tue, Oct 22, 2002 at 10:55:38PM -0700, Jason R Thorpe wrote:
> On Tue, Oct 22, 2002 at 07:43:54PM -0400, Hans-Peter Nilsson wrote:
> 
>  > For executables, make elf_object_p look at .note.ABI-tag. (They
>  > don't have the same contents on NetBSD, no?)  Create one (or
>  > something equivalent) for DSO:s too; it seems like that doesn't
>  > happen.
> 
> NetBSD uses .note.netbsd.ident ... and it's provided by the NetBSD
> crtstuff, and so only applies to linked objects (programs, shared
> libraries, etc.)
> 

.note.ABI-tag is for OS, not for linker. Linker shouldn't care
.note.ABI-tag.


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output)
  2002-10-22 16:43                                 ` Hans-Peter Nilsson
  2002-10-22 22:55                                   ` Jason R Thorpe
@ 2002-10-23  5:12                                   ` kaz Kojima
  1 sibling, 0 replies; 42+ messages in thread
From: kaz Kojima @ 2002-10-23  5:12 UTC (permalink / raw)
  To: hp; +Cc: binutils

Hans-Peter Nilsson <hp@bitrange.com> wrote:
>> Anyway, my argument about bfd_check_format_matches is for
>> "existing objects".
> 
> Um, only for objects that aren't matched by the default target
> vector, right?

Yep. I just guess that not only the default target vector but also
its associated vectors which have the opposite endian and/or the
different size could be handled specially.

Regards,
	kaz

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-14 21:38         ` H. J. Lu
@ 2002-10-14 22:05           ` Stephen Clarke
  0 siblings, 0 replies; 42+ messages in thread
From: Stephen Clarke @ 2002-10-14 22:05 UTC (permalink / raw)
  To: H. J. Lu; +Cc: binutils

From: "H. J. Lu"
Sent: Monday, October 14, 2002 9:38 PM
> 
> Now I got
> 
> FAIL: ld-sh/tlsbin-0-dso
> FAIL: ld-sh/tlsbin-1
> FAIL: ld-sh/tlsbin-2
> FAIL: ld-sh/tlsbin-3
> FAIL: ld-sh/tlsbin-4
> FAIL: ld-sh/tlspic-1
> FAIL: ld-sh/tlspic-2
> FAIL: ld-sh/tlspic-3
> FAIL: ld-sh/tlspic-4

I get these too.

> [long list of further failures]

I don't get any of these.  I'll try to work out why.
It could help if you'd send me the gas.log file.

Steve.


__________________________________________________
Do you Yahoo!?
Faith Hill - Exclusive Performances, Videos & More
http://faith.yahoo.com

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-14 21:14       ` Stephen Clarke
@ 2002-10-14 21:38         ` H. J. Lu
  2002-10-14 22:05           ` Stephen Clarke
  0 siblings, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-14 21:38 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Mon, Oct 14, 2002 at 09:14:25PM -0700, Stephen Clarke wrote:
> From: "H. J. Lu"
> Sent: Wednesday, October 02, 2002 5:04 PM
> > 
> > Here are the "make check" failures for sh64-linux. Will your patches
> > fix them?
> 
> Yes.  (Sorry to take so long to reply - I wanted to verify it.)
> Those failures should all have gone away now.  However, there are a
> few new test failures for sh64-linux related to the new tls support.
> They shouldn't be hard to sort out though.

Now I got

FAIL: ld-sh/tlsbin-0-dso
FAIL: ld-sh/tlsbin-1
FAIL: ld-sh/tlsbin-2
FAIL: ld-sh/tlsbin-3
FAIL: ld-sh/tlsbin-4
FAIL: ld-sh/tlspic-1
FAIL: ld-sh/tlspic-2
FAIL: ld-sh/tlspic-3
FAIL: ld-sh/tlspic-4
FAIL: SH PIC constructs
FAIL: sh dynamic tls
FAIL: sh pic tls
FAIL: sh non-pic tls
FAIL: Assertion .abi 32.
FAIL: Assertion .abi 64.
FAIL: Basic SHmedia instructions.
FAIL: Case-insensitive registers and opcodes.
FAIL: Case-insensitive registers and opcodes with -no-expand.
FAIL: .cranges descriptors.
FAIL: .cranges descriptors including SHcompact constant pool
FAIL: PT to SHcompact
FAIL: .cranges descriptors for SHcompact and SHmedia in .text.
FAIL: PT to SHcompact with -no-expand.
FAIL: .cranges descriptors, constant mix.
FAIL: .cranges descriptors with final variant.
FAIL: Avoid zero length .cranges range descriptor at .align in code.
FAIL: Predefined control register names.
FAIL: Predefined control register names specified in crN syntax.
FAIL: DataLabel redundant local use, SHcompact
FAIL: DataLabel redundant local use, SHmedia 32-bit ABI
FAIL: DataLabel local def/use, SHmedia 32-bit ABI
FAIL: DataLabel redundant local use, SHmedia 64-bit ABI
FAIL: DataLabel local def/use, SHmedia 64-bit ABI
FAIL: SH64 Little Endian
FAIL: SH64 Big Endian
FAIL: Immediate resolved operands, 32-bit ABI.
FAIL: Resolved 64-bit operand, 32-bit ABI.
FAIL: Immediate resolved operands, 64-bit ABI.
FAIL: Resolved 64-bit operand, 64-bit ABI.
FAIL: Dwarf2 line numbers vs macro opcodes
FAIL: Datalabel on local comm symbol and equated local comm symbol
FAIL: Mixed-ISA objects.
FAIL: Mixed-ISA objects with -no-expand.
FAIL: Assembler PC-rel resolved negative MOVI.
FAIL: MOVI expansion, 32-bit ABI.
FAIL: MOVI expansion of local symbols with relocs, 32-bit ABI.
FAIL: MOVI non-expansion of local symbols with relocs, 32-bit ABI.
FAIL: MOVI expansion, 64-bit ABI, 32-bit subset.
FAIL: MOVI expansion, 64-bit ABI, 64-bit subset.
FAIL: MOVI expansion of local symbols with relocs, 64-bit ABI.
FAIL: MOVI non-expansion of local symbols with relocs, 64-bit ABI.
FAIL: Basic SHmedia PT and PTA instructions.
FAIL: Basic SHmedia PT and PTA instructions with -no-expand.
FAIL: Inter-segment PT, 32-bit.
FAIL: Inter-segment PT, 32-bit with -no-expand.
FAIL: Inter-segment PT, 64-bit.
FAIL: Basic SHmedia PT and PTA instructions with -expand-pt32.
FAIL: Inter-segment PT, 64-bit with -expand-pt32.
FAIL: Inter-segment PT, 64-bit with -no-expand.
FAIL: PT constant, 32-bit ABI.
FAIL: PT constant, 32-bit ABI with -no-expand.
FAIL: PT constant, 64-bit ABI.
FAIL: PT constant, 64-bit ABI with -expand-pt32.
FAIL: PT constant, 64-bit ABI with -no-expand.
FAIL: PT, PTA, PTB expansion for external symbols, 32-bit ABI.
FAIL: PT, PTA, PTB non-expansion for external symbols, 32-bit ABI.
FAIL: PT, PTA, PTB expansion for external symbols, 64-bit ABI.
FAIL: PT, PTA, PTB expansion for external symbols, 64-bit ABI with -expand-pt32.
FAIL: PT, PTA, PTB non-expansion for external symbols, 64-bit ABI.
FAIL: MOVI: PC-relative relocs, 32-bit ABI.
FAIL: MOVI: PC+1-relative relocs, 32-bit ABI.
FAIL: MOVI: PC-relative datalabel relocs, 32-bit ABI.
FAIL: MOVI: PC+1-relative datalabel relocs, 32-bit ABI.
FAIL: MOVI: PC-relative reloc within .text, 32-bit ABI.
FAIL: MOVI: PC-relative relocs, 64-bit ABI.
FAIL: MOVI: PC+1-relative relocs, 64-bit ABI.
FAIL: MOVI: PC-relative datalabel relocs, 64-bit ABI.
FAIL: MOVI: PC+1-relative datalabel relocs, 64-bit ABI.
FAIL: MOVI: PC-relative reloc within .text, 64-bit ABI.
FAIL: Assembler PT relaxation limit, from first to second state.
FAIL: Assembler PTB relaxation limit, from first to second state.
FAIL: Assembler PC-rel MOVI relaxation limit, from first to second state.
FAIL: Shift expressions, 32-bit ABI.
FAIL: Shift expression, local but undefined symbol, 32-bit ABI.
FAIL: Shift expression, local but undefined symbol, 32-bit ABI with -no-expand.
FAIL: Shift expressions, 64-bit ABI, 32-bit subset.
FAIL: Shift expressions, 64-bit ABI, 64-bit subset.
FAIL: Shift expression, local but undefined symbol, 64-bit ABI.
FAIL: Shift expression, local but undefined symbol, 64-bit ABI with -no-expand.
FAIL: Minimum SH64 Syntax Support.
FAIL: Minimum SH64 Syntax Support - Pseudos.
FAIL: Unaligned pseudos, 32-bit ABI.
FAIL: Unaligned pseudos, 64-bit ABI.
FAIL: symver symver0
FAIL: symver symver1
FAIL: vtable inherit0
FAIL: vtable entry1

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-02 17:04     ` H. J. Lu
@ 2002-10-14 21:14       ` Stephen Clarke
  2002-10-14 21:38         ` H. J. Lu
  0 siblings, 1 reply; 42+ messages in thread
From: Stephen Clarke @ 2002-10-14 21:14 UTC (permalink / raw)
  To: H. J. Lu; +Cc: binutils

From: "H. J. Lu"
Sent: Wednesday, October 02, 2002 5:04 PM
> 
> Here are the "make check" failures for sh64-linux. Will your patches
> fix them?

Yes.  (Sorry to take so long to reply - I wanted to verify it.)
Those failures should all have gone away now.  However, there are a
few new test failures for sh64-linux related to the new tls support.
They shouldn't be hard to sort out though.

Steve.


__________________________________________________
Do You Yahoo!?
Sign up for SBC Yahoo! Dial - First Month Free
http://sbc.yahoo.com

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-11 10:51           ` Stephen Clarke
@ 2002-10-13 22:01             ` Hans-Peter Nilsson
  0 siblings, 0 replies; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-13 22:01 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Fri, 11 Oct 2002, Stephen Clarke wrote:
> Here's the revised patch.  I've removed the target
> exclusion from ld-sh/rd-sh.exp, and put one in
> ld-sh/ld-r-1-d.

ok.

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-09 21:20         ` Hans-Peter Nilsson
@ 2002-10-11 10:51           ` Stephen Clarke
  2002-10-13 22:01             ` Hans-Peter Nilsson
  0 siblings, 1 reply; 42+ messages in thread
From: Stephen Clarke @ 2002-10-11 10:51 UTC (permalink / raw)
  To: Hans-Peter Nilsson; +Cc: H. J. Lu, binutils

[-- Attachment #1: Type: text/plain, Size: 302 bytes --]

From: "Hans-Peter Nilsson" 
Sent: Wednesday, October 09, 2002 9:20 PM


> As long as the target exclusion is in each test rather than
> before the whole test-loop.

Here's the revised patch.  I've removed the target
exclusion from ld-sh/rd-sh.exp, and put one in
ld-sh/ld-r-1-d.

Ok to apply?

Steve.


[-- Attachment #2: sh64-ld-tests.gz --]
[-- Type: application/x-compressed, Size: 2307 bytes --]

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-09 13:34       ` Stephen Clarke
@ 2002-10-09 21:20         ` Hans-Peter Nilsson
  2002-10-11 10:51           ` Stephen Clarke
  0 siblings, 1 reply; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-09 21:20 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: H. J. Lu, binutils

On Wed, 9 Oct 2002, Stephen Clarke wrote:
> From: "Hans-Peter Nilsson"
> Subject: Re: [Patch] sh64: Fix gas testsuite expected output
> > --- ld/testsuite/ld-sh/sh.exp   30 Aug 2002 00:51:57 -0000

> > ! if { ([istarget sh64-*-elf]
> > !       || [istarget sh64*-*-linux*])} {

> > I'd rather you tweak the target specifiers on the tests:
> > #targets: sh-*-* shl*-*-* shb*-*-*
> > in each *.d should suffice.
>
> It turns out that only ld-r-1.d matches
> the sh64*-*-linux* target - none of the other *.d match
> any linux configuration.
>
> But I'm a bit nervous about the exact change you suggest,
> because I thought sh[1234]-linux were supported
> targets?

Right.

> How about the following instead:
> Index: ld-r-1.d

>   #target: sh*-*-elf sh*-*-linux*
> + #notarget: sh64*-*-linux*

Yes, using "notarget" (doh, missed that one) is fine of course.
As long as the target exclusion is in each test rather than
before the whole test-loop.

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-02 17:00     ` Hans-Peter Nilsson
  2002-10-02 17:12       ` Hans-Peter Nilsson
@ 2002-10-09 13:34       ` Stephen Clarke
  2002-10-09 21:20         ` Hans-Peter Nilsson
  1 sibling, 1 reply; 42+ messages in thread
From: Stephen Clarke @ 2002-10-09 13:34 UTC (permalink / raw)
  To: Hans-Peter Nilsson; +Cc: H. J. Lu, binutils

From: "Hans-Peter Nilsson"
Subject: Re: [Patch] sh64: Fix gas testsuite expected output


> Index: ld/testsuite/ld-sh/sh.exp
> ===================================================================
> RCS file: /cvs/src/src/ld/testsuite/ld-sh/sh.exp,v
> retrieving revision 1.8
> diff -u -c -3 -p -r1.8 sh.exp
> *** ld/testsuite/ld-sh/sh.exp   8 Feb 2002 06:49:55 -0000
> 1.8
> --- ld/testsuite/ld-sh/sh.exp   30 Aug 2002 00:51:57 -0000
> *************** if ![istarget sh*-*-*] {
> *** 25,31 ****
>       return
>   }
> 
> ! if [istarget sh64-*-elf] {
>       # relaxing not supported on sh64 yet.
>       return
>   }
> --- 25,32 ----
>       return
>   }
> 
> ! if { ([istarget sh64-*-elf]
> !       || [istarget sh64*-*-linux*])} {
>       # relaxing not supported on sh64 yet.
>       return
>   }
> 
> I'd rather you tweak the target specifiers on the tests:
> #targets: sh-*-* shl*-*-* shb*-*-*
> in each *.d should suffice.

It turns out that only ld-r-1.d matches
the sh64*-*-linux* target - none of the other *.d match
any linux configuration.

But I'm a bit nervous about the exact change you suggest,
because I thought sh[1234]-linux were supported
targets?

How about the following instead:
Index: ld-r-1.d
===================================================================
RCS file: /cvs/src/src/ld/testsuite/ld-sh/ld-r-1.d,v
retrieving revision 1.3
diff -c -3 -p -r1.3 ld-r-1.d
*** ld-r-1.d 24 Feb 2002 16:44:44 -0000 1.3
--- ld-r-1.d 9 Oct 2002 20:28:27 -0000
***************
*** 4,9 ****
--- 4,10 ----
  #ld: -r -EL
  #readelf: -r -x1 -x2
  #target: sh*-*-elf sh*-*-linux*
+ #notarget: sh64*-*-linux*
  
  # Make sure relocations against global and local symbols with relative and
  # absolute 32-bit relocs don't come out wrong after ld -r.  Remember that




__________________________________________________
Do you Yahoo!?
Faith Hill - Exclusive Performances, Videos & More
http://faith.yahoo.com

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-02 17:00     ` Hans-Peter Nilsson
@ 2002-10-02 17:12       ` Hans-Peter Nilsson
  2002-10-09 13:34       ` Stephen Clarke
  1 sibling, 0 replies; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-02 17:12 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: H. J. Lu, binutils

On Wed, 2 Oct 2002, Hans-Peter Nilsson wrote:
> On Wed, 2 Oct 2002, Stephen Clarke wrote:
> > http://sources.redhat.com/ml/binutils/2002-09/msg00050.html
>
> But for one thing (leading to some other changes), the patch is
> ok:
>
> Index: ld/testsuite/ld-sh/sh.exp

> I'd rather you tweak the target specifiers on the tests:
> #targets: sh-*-* shl*-*-* shb*-*-*
> in each *.d should suffice.

I quoted the wrong patch, so the above probably looked
confusing, sorry.  I meant this one:

Index: ld/testsuite/ld-sh/rd-sh.exp
===================================================================
RCS file: /cvs/src/src/ld/testsuite/ld-sh/rd-sh.exp,v
retrieving revision 1.1
diff -u -c -3 -p -r1.1 rd-sh.exp
*** ld/testsuite/ld-sh/rd-sh.exp        14 Oct 2001 15:18:11 -0000      1.1
--- ld/testsuite/ld-sh/rd-sh.exp        30 Aug 2002 00:51:57 -0000
***************
*** 27,33 ****
  if ![istarget sh*-*-*] {
      return
  }
!
  set rd_test_list [lsort [glob -nocomplain $srcdir/$subdir/*.d]]
  foreach shtest $rd_test_list {
      # We need to strip the ".d", but can leave the dirname.
--- 27,35 ----
  if ![istarget sh*-*-*] {
      return
  }
! if [istarget sh64*-*-linux*] {
!     return
! }
  set rd_test_list [lsort [glob -nocomplain $srcdir/$subdir/*.d]]
  foreach shtest $rd_test_list {
      # We need to strip the ".d", but can leave the dirname.

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-02 16:28   ` Stephen Clarke
  2002-10-02 17:00     ` Hans-Peter Nilsson
@ 2002-10-02 17:04     ` H. J. Lu
  2002-10-14 21:14       ` Stephen Clarke
  1 sibling, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-02 17:04 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Wed, Oct 02, 2002 at 04:22:08PM -0700, Stephen Clarke wrote:
> From: "H. J. Lu"
> Sent: Wednesday, October 02, 2002 4:08 PM
> 
> 
> > Is sh64-linux a supported Linux target? If yes, I can add it to my
> > binutils test target. I already test sh-linux, adding sh64-linux won't
> > be a problem.
> 
> Not yet ... there's a patch at
> http://sources.redhat.com/ml/binutils/2002-09/msg00048.html
> waiting to be reviewed.
> 
> This would need sh64 gas and ld testsuite fixes too, see
> http://sources.redhat.com/ml/binutils/2002-09/msg00049.html
> http://sources.redhat.com/ml/binutils/2002-09/msg00050.html
> 
> I couldn't miss this opportunity to nag the maintainers could I :-)
> 

Here are the "make check" failures for sh64-linux. Will your patches
fix them?


H.J.
-----
build-sh64-linux/ld/ld.log:FAIL: ld-discard/extern
build-sh64-linux/ld/ld.log:FAIL: ld-discard/start
build-sh64-linux/ld/ld.log:FAIL: ld-discard/static
build-sh64-linux/ld/ld.log:FAIL: dynamic sections
build-sh64-linux/ld/ld.log:FAIL: overlay size
build-sh64-linux/ld/ld.log:FAIL: overlay size (map check)
build-sh64-linux/ld/ld.log:FAIL: Preserve default . = 0
build-sh64-linux/ld/ld.log:FAIL: Preserve explicit . = 0
build-sh64-linux/ld/ld.log:FAIL: ld-sh/ld-r-1
build-sh64-linux/ld/ld.log:FAIL: SH relaxing to S-records
build-sh64-linux/gas/testsuite/gas.log:FAIL: pcrel values in assignment
build-sh64-linux/gas/testsuite/gas.log:FAIL: simplifiable double subtraction
build-sh64-linux/gas/testsuite/gas.log:FAIL: simplifiable double subtraction (-a)
build-sh64-linux/gas/testsuite/gas.log:FAIL: simple FP constants
build-sh64-linux/gas/testsuite/gas.log:FAIL: comment.s: comments in listings
build-sh64-linux/gas/testsuite/gas.log:FAIL: difference between forward references
build-sh64-linux/gas/testsuite/gas.log:FAIL: struct
build-sh64-linux/gas/testsuite/gas.log:FAIL: align
build-sh64-linux/gas/testsuite/gas.log:FAIL: align2
build-sh64-linux/gas/testsuite/gas.log:FAIL: conditional listings
build-sh64-linux/gas/testsuite/gas.log:FAIL: incbin
build-sh64-linux/gas/testsuite/gas.log:FAIL: elf ehopt0
build-sh64-linux/gas/testsuite/gas.log:FAIL: elf section0
build-sh64-linux/gas/testsuite/gas.log:FAIL: elf section1
build-sh64-linux/gas/testsuite/gas.log:FAIL: elf section2 list
build-sh64-linux/gas/testsuite/gas.log:FAIL: macro test 1
build-sh64-linux/gas/testsuite/gas.log:FAIL: macro test 2
build-sh64-linux/gas/testsuite/gas.log:FAIL: macro test 3
build-sh64-linux/gas/testsuite/gas.log:FAIL: macro irp
build-sh64-linux/gas/testsuite/gas.log:FAIL: macro rept
build-sh64-linux/gas/testsuite/gas.log:FAIL: semi
build-sh64-linux/gas/testsuite/gas.log:FAIL: strings
build-sh64-linux/gas/testsuite/gas.log:FAIL: fp.s: floating point tests (sh3e)
build-sh64-linux/gas/testsuite/gas.log:FAIL: SH DSP basic instructions
build-sh64-linux/gas/testsuite/gas.log:FAIL: PC-relative loads
build-sh64-linux/gas/testsuite/gas.log:FAIL: PC-relative loads
build-sh64-linux/gas/testsuite/gas.log:FAIL: SH PIC constructs
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 6)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 7)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 8)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 9)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 10)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 11)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 12)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 13)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 14)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 15)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 16)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 17)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 18)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 19)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 20)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 21)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 22)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 23)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 24)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 25)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 26)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 27)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 28)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 29)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 30)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 31)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 32)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 33)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 34)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 35)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 36)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 37)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 38)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 39)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 40)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 41)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 42)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 43)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 44)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 45)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 46)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s  (test for errors, line 47)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-1.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-2.s  (test for errors, line 9)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-2.s  (test for errors, line 10)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-2.s  (test for errors, line 13)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-2.s  (test for errors, line 14)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-2.s  (test for errors, line 15)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-2.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-3.s  (test for errors, line 12)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-3.s  (test for errors, line 16)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-3.s  (test for errors, line 18)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-3.s  (test for errors, line 20)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-3.s  (test for errors, line 21)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-3.s  (test for errors, line 22)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-3.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-4.s  (test for errors, line 12)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-4.s  (test for errors, line 19)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-4.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-abi-32.s  (test for errors, line 7)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-abi-32.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-abi-64.s  (test for errors, line 7)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-abi-64.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-dsp.s  (test for errors, line 13)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-dsp.s  (test for errors, line 14)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-dsp.s  (test for errors, line 15)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-dsp.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-movi-noexp-1.s  (test for errors, line 9)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-movi-noexp-1.s  (test for errors, line 10)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-movi-noexp-1.s  (test for errors, line 11)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-movi-noexp-1.s  (test for errors, line 13)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-movi-noexp-1.s  (test for errors, line 14)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-movi-noexp-1.s  (test for errors, line 16)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-movi-noexp-1.s  (test for errors, line 17)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-movi-noexp-1.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-noexp-cmd1.s  (test for errors, line )
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-noexp-cmd1.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt-1.s  (test for errors, line 10)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt-1.s  (test for errors, line 17)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt-1.s  (test for errors, line 18)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt-1.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt32-cmd1.s  (test for errors, line )
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt32-cmd1.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt32-cmd2.s  (test for errors, line )
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt32-cmd2.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt32-cmd3.s  (test for errors, line )
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-pt32-cmd3.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-1.s  (test for errors, line 10)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-1.s  (test for errors, line 13)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-1.s  (test for errors, line 15)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-1.s  (test for errors, line 33)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-1.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-2.s  (test for errors, line 10)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-2.s  (test for errors, line 13)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-2.s  (test for errors, line 15)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-2.s  (test for errors, line 33)
build-sh64-linux/gas/testsuite/gas.log:FAIL: gas/sh/sh64/err-ptb-2.s (test for excess errors)
build-sh64-linux/gas/testsuite/gas.log:FAIL: Assertion .abi 32.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Assertion .abi 64.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Basic SHmedia instructions.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Case-insensitive registers and opcodes.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Case-insensitive registers and opcodes with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: .cranges descriptors.
build-sh64-linux/gas/testsuite/gas.log:FAIL: .cranges descriptors including SHcompact constant pool
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT to SHcompact
build-sh64-linux/gas/testsuite/gas.log:FAIL: .cranges descriptors for SHcompact and SHmedia in .text.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT to SHcompact with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: .cranges descriptors, constant mix.
build-sh64-linux/gas/testsuite/gas.log:FAIL: .cranges descriptors with final variant.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Avoid zero length .cranges range descriptor at .align in code.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Predefined control register names.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Predefined control register names specified in crN syntax.
build-sh64-linux/gas/testsuite/gas.log:FAIL: DataLabel redundant local use, SHcompact
build-sh64-linux/gas/testsuite/gas.log:FAIL: DataLabel redundant local use, SHmedia 32-bit ABI
build-sh64-linux/gas/testsuite/gas.log:FAIL: DataLabel local def/use, SHmedia 32-bit ABI
build-sh64-linux/gas/testsuite/gas.log:FAIL: DataLabel redundant local use, SHmedia 64-bit ABI
build-sh64-linux/gas/testsuite/gas.log:FAIL: DataLabel local def/use, SHmedia 64-bit ABI
build-sh64-linux/gas/testsuite/gas.log:FAIL: SH64 Little Endian
build-sh64-linux/gas/testsuite/gas.log:FAIL: SH64 Big Endian
build-sh64-linux/gas/testsuite/gas.log:FAIL: Immediate resolved operands, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Resolved 64-bit operand, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Immediate resolved operands, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Resolved 64-bit operand, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Dwarf2 line numbers vs macro opcodes
build-sh64-linux/gas/testsuite/gas.log:FAIL: Datalabel on local comm symbol and equated local comm symbol
build-sh64-linux/gas/testsuite/gas.log:FAIL: Mixed-ISA objects.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Mixed-ISA objects with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Assembler PC-rel resolved negative MOVI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI expansion, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI expansion of local symbols with relocs, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI non-expansion of local symbols with relocs, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI expansion, 64-bit ABI, 32-bit subset.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI expansion, 64-bit ABI, 64-bit subset.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI expansion of local symbols with relocs, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI non-expansion of local symbols with relocs, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Basic SHmedia PT and PTA instructions.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Basic SHmedia PT and PTA instructions with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Inter-segment PT, 32-bit.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Inter-segment PT, 32-bit with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Inter-segment PT, 64-bit.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Basic SHmedia PT and PTA instructions with -expand-pt32.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Inter-segment PT, 64-bit with -expand-pt32.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Inter-segment PT, 64-bit with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT constant, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT constant, 32-bit ABI with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT constant, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT constant, 64-bit ABI with -expand-pt32.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT constant, 64-bit ABI with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT, PTA, PTB expansion for external symbols, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT, PTA, PTB non-expansion for external symbols, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT, PTA, PTB expansion for external symbols, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT, PTA, PTB expansion for external symbols, 64-bit ABI with -expand-pt32.
build-sh64-linux/gas/testsuite/gas.log:FAIL: PT, PTA, PTB non-expansion for external symbols, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC-relative relocs, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC+1-relative relocs, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC-relative datalabel relocs, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC+1-relative datalabel relocs, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC-relative reloc within .text, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC-relative relocs, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC+1-relative relocs, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC-relative datalabel relocs, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC+1-relative datalabel relocs, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: MOVI: PC-relative reloc within .text, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Assembler PT relaxation limit, from first to second state.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Assembler PTB relaxation limit, from first to second state.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Assembler PC-rel MOVI relaxation limit, from first to second state.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Shift expressions, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Shift expression, local but undefined symbol, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Shift expression, local but undefined symbol, 32-bit ABI with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Shift expressions, 64-bit ABI, 32-bit subset.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Shift expressions, 64-bit ABI, 64-bit subset.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Shift expression, local but undefined symbol, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Shift expression, local but undefined symbol, 64-bit ABI with -no-expand.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Minimum SH64 Syntax Support.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Minimum SH64 Syntax Support - Pseudos.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Unaligned pseudos, 32-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: Unaligned pseudos, 64-bit ABI.
build-sh64-linux/gas/testsuite/gas.log:FAIL: symver symver0
build-sh64-linux/gas/testsuite/gas.log:FAIL: symver symver1
build-sh64-linux/gas/testsuite/gas.log:FAIL: symver symver2
build-sh64-linux/gas/testsuite/gas.log:FAIL: symver symver3
build-sh64-linux/gas/testsuite/gas.log:FAIL: symver symver6
build-sh64-linux/gas/testsuite/gas.log:FAIL: vtable inherit0
build-sh64-linux/gas/testsuite/gas.log:FAIL: vtable inherit1
build-sh64-linux/gas/testsuite/gas.log:FAIL: vtable entry1

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-02 16:28   ` Stephen Clarke
@ 2002-10-02 17:00     ` Hans-Peter Nilsson
  2002-10-02 17:12       ` Hans-Peter Nilsson
  2002-10-09 13:34       ` Stephen Clarke
  2002-10-02 17:04     ` H. J. Lu
  1 sibling, 2 replies; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-02 17:00 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: H. J. Lu, binutils

On Wed, 2 Oct 2002, Stephen Clarke wrote:

While I have your attention, can you please do some official
check regarding the TLS ELF reloc numbers that Kojima-san
proposed?

> From: "H. J. Lu"
> Sent: Wednesday, October 02, 2002 4:08 PM
>
>
> > Is sh64-linux a supported Linux target? If yes, I can add it to my
> > binutils test target. I already test sh-linux, adding sh64-linux won't
> > be a problem.
>
> Not yet ... there's a patch at
> http://sources.redhat.com/ml/binutils/2002-09/msg00048.html
> waiting to be reviewed.

That patch is ok.

> This would need sh64 gas and ld testsuite fixes too, see
> http://sources.redhat.com/ml/binutils/2002-09/msg00049.html

Ok.

> http://sources.redhat.com/ml/binutils/2002-09/msg00050.html

But for one thing (leading to some other changes), the patch is
ok:

Index: ld/testsuite/ld-sh/sh.exp
===================================================================
RCS file: /cvs/src/src/ld/testsuite/ld-sh/sh.exp,v
retrieving revision 1.8
diff -u -c -3 -p -r1.8 sh.exp
*** ld/testsuite/ld-sh/sh.exp   8 Feb 2002 06:49:55 -0000
1.8
--- ld/testsuite/ld-sh/sh.exp   30 Aug 2002 00:51:57 -0000
*************** if ![istarget sh*-*-*] {
*** 25,31 ****
      return
  }

! if [istarget sh64-*-elf] {
      # relaxing not supported on sh64 yet.
      return
  }
--- 25,32 ----
      return
  }

! if { ([istarget sh64-*-elf]
!       || [istarget sh64*-*-linux*])} {
      # relaxing not supported on sh64 yet.
      return
  }

I'd rather you tweak the target specifiers on the tests:
#targets: sh-*-* shl*-*-* shb*-*-*
in each *.d should suffice.

>
> I couldn't miss this opportunity to nag the maintainers could I :-)

Where's Joern these days?  Doesn't he work on SH stuff?

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-02 16:08 ` H. J. Lu
@ 2002-10-02 16:28   ` Stephen Clarke
  2002-10-02 17:00     ` Hans-Peter Nilsson
  2002-10-02 17:04     ` H. J. Lu
  0 siblings, 2 replies; 42+ messages in thread
From: Stephen Clarke @ 2002-10-02 16:28 UTC (permalink / raw)
  To: H. J. Lu; +Cc: binutils

From: "H. J. Lu"
Sent: Wednesday, October 02, 2002 4:08 PM


> Is sh64-linux a supported Linux target? If yes, I can add it to my
> binutils test target. I already test sh-linux, adding sh64-linux won't
> be a problem.

Not yet ... there's a patch at
http://sources.redhat.com/ml/binutils/2002-09/msg00048.html
waiting to be reviewed.

This would need sh64 gas and ld testsuite fixes too, see
http://sources.redhat.com/ml/binutils/2002-09/msg00049.html
http://sources.redhat.com/ml/binutils/2002-09/msg00050.html

I couldn't miss this opportunity to nag the maintainers could I :-)

Steve.


__________________________________________________
Do You Yahoo!?
Sign up for SBC Yahoo! Dial - First Month Free
http://sbc.yahoo.com

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-02 16:00 Stephen Clarke
  2002-10-02 16:05 ` Hans-Peter Nilsson
@ 2002-10-02 16:08 ` H. J. Lu
  2002-10-02 16:28   ` Stephen Clarke
  1 sibling, 1 reply; 42+ messages in thread
From: H. J. Lu @ 2002-10-02 16:08 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Wed, Oct 02, 2002 at 02:40:10PM -0700, Stephen Clarke wrote:
> The attached file contains a patch to correct the objdump
> expected output for two tests in the sh64 gas testsuite.
> 

Is sh64-linux a supported Linux target? If yes, I can add it to my
binutils test target. I already test sh-linux, adding sh64-linux won't
be a problem.


H.J.

^ permalink raw reply	[flat|nested] 42+ messages in thread

* Re: [Patch] sh64: Fix gas testsuite expected output
  2002-10-02 16:00 Stephen Clarke
@ 2002-10-02 16:05 ` Hans-Peter Nilsson
  2002-10-02 16:08 ` H. J. Lu
  1 sibling, 0 replies; 42+ messages in thread
From: Hans-Peter Nilsson @ 2002-10-02 16:05 UTC (permalink / raw)
  To: Stephen Clarke; +Cc: binutils

On Wed, 2 Oct 2002, Stephen Clarke wrote:
> The change is because the *ABS* symbol is no longer
> being generated for these tests, I think this is probably a
> result of the stage 1 gas relocs rewrite (though this
> is the first time I've checked the sh64 testsuite for about
> a month).

Eventually Alan got that silly symbol out, yes. :-)

> Ok to apply?

Yes, thanks.

brgds, H-P

^ permalink raw reply	[flat|nested] 42+ messages in thread

* [Patch] sh64: Fix gas testsuite expected output
@ 2002-10-02 16:00 Stephen Clarke
  2002-10-02 16:05 ` Hans-Peter Nilsson
  2002-10-02 16:08 ` H. J. Lu
  0 siblings, 2 replies; 42+ messages in thread
From: Stephen Clarke @ 2002-10-02 16:00 UTC (permalink / raw)
  To: binutils

[-- Attachment #1: Type: text/plain, Size: 374 bytes --]

The attached file contains a patch to correct the objdump
expected output for two tests in the sh64 gas testsuite.

The change is because the *ABS* symbol is no longer
being generated for these tests, I think this is probably a
result of the stage 1 gas relocs rewrite (though this
is the first time I've checked the sh64 testsuite for about
a month).

Ok to apply?
Steve.


[-- Attachment #2: gastests-patch.txt --]
[-- Type: text/plain, Size: 1692 bytes --]

2002-10-02  Stephen Clarke <stephen.clarke@superh.com>

	* gas/sh/sh64/ptc32-noexp-1.d: Adjust for changes to *ABS*
	symbol generation.
	* gas/sh/sh64/ptc64-noexp-1.d: Likewise.
	
Index: gas/testsuite/gas/sh/sh64/ptc32-noexp-1.d
===================================================================
RCS file: /cvs/src/src/gas/testsuite/gas/sh/sh64/ptc32-noexp-1.d,v
retrieving revision 1.1
diff -u -c -3 -p -r1.1 ptc32-noexp-1.d
*** gas/testsuite/gas/sh/sh64/ptc32-noexp-1.d	8 Feb 2002 06:46:14 -0000	1.1
--- gas/testsuite/gas/sh/sh64/ptc32-noexp-1.d	2 Oct 2002 21:27:39 -0000
***************
*** 8,12 ****
  Disassembly of section \.text:
  
  [0]+ <start>:
! [ 	]+0:[ 	]+e8000610[ 	]+pta/l	4 <\*ABS\*\+0x4>,tr1
  [ 	]+0:[ 	]+R_SH_PT_16	\*ABS\*\+0x100
--- 8,12 ----
  Disassembly of section \.text:
  
  [0]+ <start>:
! [ 	]+0:[ 	]+e8000610[ 	]+pta/l	4 <start\+0x4>,tr1
  [ 	]+0:[ 	]+R_SH_PT_16	\*ABS\*\+0x100
Index: gas/testsuite/gas/sh/sh64/ptc64-noexp-1.d
===================================================================
RCS file: /cvs/src/src/gas/testsuite/gas/sh/sh64/ptc64-noexp-1.d,v
retrieving revision 1.1
diff -u -c -3 -p -r1.1 ptc64-noexp-1.d
*** gas/testsuite/gas/sh/sh64/ptc64-noexp-1.d	8 Feb 2002 06:46:14 -0000	1.1
--- gas/testsuite/gas/sh/sh64/ptc64-noexp-1.d	2 Oct 2002 21:27:39 -0000
***************
*** 8,13 ****
  Disassembly of section \.text:
  
  [0]+ <start>:
! [ 	]+0:[ 	]+e8000610[ 	]+pta/l	4 <\*ABS\*\+0x4>,tr1
  [ 	]+0:[ 	]+R_SH_PT_16	\*ABS\*\+0x100
  
--- 8,13 ----
  Disassembly of section \.text:
  
  [0]+ <start>:
! [ 	]+0:[ 	]+e8000610[ 	]+pta/l	4 <start\+0x4>,tr1
  [ 	]+0:[ 	]+R_SH_PT_16	\*ABS\*\+0x100
  

^ permalink raw reply	[flat|nested] 42+ messages in thread

end of thread, other threads:[~2002-10-23 12:12 UTC | newest]

Thread overview: 42+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2002-10-15 20:16 [Patch] sh64: Fix gas testsuite expected output Stephen Clarke
2002-10-15 20:27 ` H. J. Lu
2002-10-15 21:39   ` Stephen Clarke
2002-10-15 22:32     ` H. J. Lu
2002-10-15 22:54       ` Stephen Clarke
2002-10-15 23:00         ` H. J. Lu
2002-10-15 23:39           ` H. J. Lu
2002-10-16  1:47             ` Alan Modra
2002-10-16  6:16               ` H. J. Lu
2002-10-16  8:03                 ` Alan Modra
2002-10-16  8:14                   ` H. J. Lu
2002-10-16  9:50                   ` Matt Thomas
2002-10-18  1:56                     ` Alan Modra
2002-10-18 10:22                       ` Ulrich Drepper
2002-10-18 10:38                         ` H. J. Lu
2002-10-18 11:21                           ` Jason R Thorpe
2002-10-18 11:27                             ` Jason R Thorpe
2002-10-18 12:33                         ` Hans-Peter Nilsson
2002-10-18 18:04                           ` Alan Modra
2002-10-18 18:46                             ` Hans-Peter Nilsson
2002-10-22  1:44                           ` kaz Kojima
2002-10-22 10:14                             ` Change ELFOSABI in SH (Re: [Patch] sh64: Fix gas testsuite expected output) H. J. Lu
2002-10-22 15:58                               ` kaz Kojima
2002-10-22 16:43                                 ` Hans-Peter Nilsson
2002-10-22 22:55                                   ` Jason R Thorpe
2002-10-22 22:58                                     ` H. J. Lu
2002-10-23  5:12                                   ` kaz Kojima
2002-10-15 21:51 ` [Patch] sh64: Fix gas testsuite expected output Hans-Peter Nilsson
  -- strict thread matches above, loose matches on Subject: below --
2002-10-02 16:00 Stephen Clarke
2002-10-02 16:05 ` Hans-Peter Nilsson
2002-10-02 16:08 ` H. J. Lu
2002-10-02 16:28   ` Stephen Clarke
2002-10-02 17:00     ` Hans-Peter Nilsson
2002-10-02 17:12       ` Hans-Peter Nilsson
2002-10-09 13:34       ` Stephen Clarke
2002-10-09 21:20         ` Hans-Peter Nilsson
2002-10-11 10:51           ` Stephen Clarke
2002-10-13 22:01             ` Hans-Peter Nilsson
2002-10-02 17:04     ` H. J. Lu
2002-10-14 21:14       ` Stephen Clarke
2002-10-14 21:38         ` H. J. Lu
2002-10-14 22:05           ` Stephen Clarke

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).