More MIPS testsuite fixes

More MIPS testsuite fixes

[Daniel Jacobowitz]

hidden2 needs to use .type, because MIPS defaults to OBJECT.  It also needs
to place .dynamic; this is because the PT_DYNAMIC segment on MIPS includes
.dynsym, and orphan placement will happen to place .dynamic after .dynsym,
which ld complains about.

ld-elf/start.s needs to provide __start since that's the default entry point
on MIPS.

warn1.d needs to use a linker script without SIZEOF_HEADERS, because
mips-linux will have a SEC_ALLOC reginfo section, which the linker script
places up after 0x400000.  So we run out of program headers.  Using group.ld
should handle the Alpha case as well as -Ttext did.  Alan, I gather you have
a list of targets that you run through the testsuite for; could you add
mips-linux?

Is the patch OK?

-- 
Daniel Jacobowitz
CodeSourcery, LLC

2005-02-21  Daniel Jacobowitz  <dan@codesourcery.com>

	* ld-elfvsb/hidden2.d: Expect OBJECT.
	* ld-elfvsb/hidden2.s: Force type OBJECT.
	* ld-elfvsb/hidden2.ld: Place .dynamic explicitly.
	* ld-elf/start.s: Provide __start for MIPS.
	* ld-elfcomm/elfcomm.exp: Accept MIPS common section.  XFAIL size change
	test.
	* ld-elf/warn1.d: Use group.ld instead of -Ttext.

Index: binutils/ld/testsuite/ld-elfvsb/hidden2.d
===================================================================
--- binutils.orig/ld/testsuite/ld-elfvsb/hidden2.d	2005-02-10 09:09:43.000000000 -0500
+++ binutils/ld/testsuite/ld-elfvsb/hidden2.d	2005-02-21 09:38:31.167154608 -0500
@@ -5,5 +5,5 @@
 
 Symbol table for image:
 #...
-[ 	]*[0-9]+ +[0-9]+: [0-9a-fA-F]* +0  NOTYPE  LOCAL HIDDEN +ABS foo
+[ 	]*[0-9]+ +[0-9]+: [0-9a-fA-F]* +0  OBJECT  LOCAL HIDDEN +ABS foo
 #pass
Index: binutils/ld/testsuite/ld-elfvsb/hidden2.ld
===================================================================
--- binutils.orig/ld/testsuite/ld-elfvsb/hidden2.ld	2005-02-14 11:54:32.000000000 -0500
+++ binutils/ld/testsuite/ld-elfvsb/hidden2.ld	2005-02-21 09:38:31.167154608 -0500
@@ -3,4 +3,5 @@ SECTIONS
   . = 0x1000;
   PROVIDE (foo = .);
   .data : { *(.data) }
+  .dynamic : { *(.dynamic) }
 }
Index: binutils/ld/testsuite/ld-elfvsb/hidden2.s
===================================================================
--- binutils.orig/ld/testsuite/ld-elfvsb/hidden2.s	2005-02-21 09:37:18.221944952 -0500
+++ binutils/ld/testsuite/ld-elfvsb/hidden2.s	2005-02-21 09:38:49.649752748 -0500
@@ -1,4 +1,5 @@
 	.data
 	.hidden foo
 	.global foo
+	.type foo,%object
 	.dc.a foo
Index: binutils/ld/testsuite/ld-elf/start.s
===================================================================
--- binutils.orig/ld/testsuite/ld-elf/start.s	2005-02-06 21:46:15.000000000 -0500
+++ binutils/ld/testsuite/ld-elf/start.s	2005-02-21 09:38:31.168154425 -0500
@@ -1,4 +1,6 @@
 	.text
 	.global _start
 _start:
+	.global __start
+__start:
 	.long 0
Index: binutils/ld/testsuite/ld-elfcomm/elfcomm.exp
===================================================================
--- binutils.orig/ld/testsuite/ld-elfcomm/elfcomm.exp	2003-06-17 23:38:05.000000000 -0400
+++ binutils/ld/testsuite/ld-elfcomm/elfcomm.exp	2005-02-21 09:38:31.168154425 -0500
@@ -45,7 +45,7 @@ proc dump_common1 { testname } {
 
     send_log "$READELF -s tmpdir/common1.o | grep foo\n"
     catch "exec $READELF -s tmpdir/common1.o | grep foo" exec_output
-    if { ![regexp "(\[ 	\]*)(\[0-9\]+):(\[ 	\]*)(\[0\]*)80(\[ 	\]+)4(\[ 	\]+)OBJECT(\[ 	\]+)GLOBAL(\[ 	\]+)DEFAULT(\[ 	\]+)COM(\[ 	\]+)_?foo2" $exec_output]
+    if { ![regexp "(\[ 	\]*)(\[0-9\]+):(\[ 	\]*)(\[0\]*)80(\[ 	\]+)4(\[ 	\]+)OBJECT(\[ 	\]+)GLOBAL(\[ 	\]+)DEFAULT(\[ 	\]+)(PRC\\\[0xff03\\\]|COM)(\[ 	\]+)_?foo2" $exec_output]
  	 || ![regexp "(\[ 	\]*)(\[0-9\]+):(\[ 	\]*)(\[0-9\]+)(\[ 	\]+)21(\[ 	\]+)OBJECT(\[ 	\]+)GLOBAL(\[ 	\]+)DEFAULT(\[ 	\]+)(\[0-9\]+)(\[ 	\]+)_?foo1" $exec_output] } {
 	send_log "$exec_output\n"
 	verbose $exec_output
@@ -72,7 +72,13 @@ if { [ld_simple_link $ld tmpdir/common1.
 
 if { ![regexp "Warning: alignment (\[0-9\]+) of symbol \`_?foo1\' in tmpdir/common1b.o is smaller than 64 in tmpdir/common1a.o" $link_output]
      || ![regexp "Warning: size of symbol \`_?foo1\' changed from 2 in tmpdir/common1a.o to 21 in tmpdir/common1b.o" $link_output] } {
-    fail $test1w1
+    if { [istarget mips*-*-*] } {
+	# This test fails on MIPS because the backend sets type_change_ok.  The
+	# size change warning is suppressed.
+	xfail $test1w1
+    } else {
+	fail $test1w1
+    }
 } else {
     pass $test1w1
 }
Index: binutils/ld/testsuite/ld-elf/warn1.d
===================================================================
--- binutils.orig/ld/testsuite/ld-elf/warn1.d	2005-02-21 09:37:17.218112989 -0500
+++ binutils/ld/testsuite/ld-elf/warn1.d	2005-02-21 10:39:19.599477177 -0500
@@ -1,7 +1,7 @@
 #source: start.s
 #source: symbol1ref.s
 #source: symbol1w.s
-#ld: -Ttext 0x1000
+#ld: -T group.ld
 #warning: ^[^\\n]*\): warning: witty one-liner$
 #readelf: -s
 #notarget: "sparc64-*-solaris2*" "sparcv9-*-solaris2*"

Re: More MIPS testsuite fixes

[Alan Modra]

On Mon, Feb 21, 2005 at 10:50:57AM -0500, Daniel Jacobowitz wrote:
> Alan, I gather you have
> a list of targets that you run through the testsuite for; could you add
> mips-linux?

I already have it on my list (and had seen the failures, just not fixed
them).  Hmm, since we are talking mips failures, I also test mips-ecoff:

$ grep FAIL check.log0 | wc -l
    318

How about someone in the mips camp fixing them, or deprecate the target?

> Is the patch OK?

Yes, looks good.

-- 
Alan Modra
IBM OzLabs - Linux Technology Centre

Re: More MIPS testsuite fixes

[Daniel Jacobowitz]

On Tue, Feb 22, 2005 at 09:36:21AM +1030, Alan Modra wrote:
> On Mon, Feb 21, 2005 at 10:50:57AM -0500, Daniel Jacobowitz wrote:
> > Alan, I gather you have
> > a list of targets that you run through the testsuite for; could you add
> > mips-linux?
> 
> I already have it on my list (and had seen the failures, just not fixed
> them).

Oh, I see why I guessed it wasn't - I'd been sitting on this patch for
a couple of days.  The test was broken for mips-linux before and after
your -Ttext change, but -Ttext broke it with the previous patch I had
in my working directory.  Sorry.

>  Hmm, since we are talking mips failures, I also test mips-ecoff:
> 
> $ grep FAIL check.log0 | wc -l
>     318
> 
> How about someone in the mips camp fixing them, or deprecate the target?

We'd been talking about deprecating it.  I don't remember the result of
the discussion.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: More MIPS testsuite fixes

[Thiemo Seufer]

Daniel Jacobowitz wrote:
[snip]
> >  Hmm, since we are talking mips failures, I also test mips-ecoff:
> > 
> > $ grep FAIL check.log0 | wc -l
> >     318
> > 
> > How about someone in the mips camp fixing them, or deprecate the target?

A while ago most of those were ELF-only testcases which are erraneously
run for mips-ecoff.

> We'd been talking about deprecating it.  I don't remember the result of
> the discussion.

It is still in use for bootloader related stuff on Linux and NetBSD.


Thiemo

Re: More MIPS testsuite fixes

[Daniel Jacobowitz]

On Tue, Feb 22, 2005 at 01:43:16AM +0100, Thiemo Seufer wrote:
> Daniel Jacobowitz wrote:
> [snip]
> > >  Hmm, since we are talking mips failures, I also test mips-ecoff:
> > > 
> > > $ grep FAIL check.log0 | wc -l
> > >     318
> > > 
> > > How about someone in the mips camp fixing them, or deprecate the target?
> 
> A while ago most of those were ELF-only testcases which are erraneously
> run for mips-ecoff.
> 
> > We'd been talking about deprecating it.  I don't remember the result of
> > the discussion.
> 
> It is still in use for bootloader related stuff on Linux and NetBSD.

That can't be resolved by using ELF and objcopy?  That's where the
discussion got stalled last time IIRC.

-- 
Daniel Jacobowitz
CodeSourcery, LLC

Re: More MIPS testsuite fixes

[Eric Christopher]

On Tue, 2005-02-22 at 13:29 -0500, Daniel Jacobowitz wrote:
> On Tue, Feb 22, 2005 at 01:43:16AM +0100, Thiemo Seufer wrote:
> > Daniel Jacobowitz wrote:
> > [snip]
> > > >  Hmm, since we are talking mips failures, I also test mips-ecoff:
> > > > 
> > > > $ grep FAIL check.log0 | wc -l
> > > >     318
> > > > 
> > > > How about someone in the mips camp fixing them, or deprecate the target?
> > 
> > A while ago most of those were ELF-only testcases which are erraneously
> > run for mips-ecoff.
> > 
> > > We'd been talking about deprecating it.  I don't remember the result of
> > > the discussion.
> > 
> > It is still in use for bootloader related stuff on Linux and NetBSD.
> 
> That can't be resolved by using ELF and objcopy?  That's where the
> discussion got stalled last time IIRC.
> 

I talked to some of the NetBSD guys yesterday. They believe that they
have their own elf to ecoff converter and don't need it, but I sent mail
to their toolchain list for verification. I'll take silence of a couple
of days as assent. :)

-eric

Re: More MIPS testsuite fixes

[Maciej W. Rozycki]

On Tue, 22 Feb 2005, Eric Christopher wrote:

> > That can't be resolved by using ELF and objcopy?  That's where the
> > discussion got stalled last time IIRC.
> > 
> 
> I talked to some of the NetBSD guys yesterday. They believe that they
> have their own elf to ecoff converter and don't need it, but I sent mail
> to their toolchain list for verification. I'll take silence of a couple
> of days as assent. :)

 I'd keep stuff like objcopy and objdump working to be able to do 
conversions and verify they have actually worked.  Otherwise I don't care 
about ECOFF either.

 But this is not the right mailing list for people that actually use that 
stuff, so it may be quite a good idea to ask at a more appropriate forum, 
like <linux-mips@linux-mips.org>.

 Also people may still keep using DEC Ultrix somewhere. ;-)

  Maciej

Re: More MIPS testsuite fixes

[Eric Christopher]

>  I'd keep stuff like objcopy and objdump working to be able to do 
> conversions and verify they have actually worked.  Otherwise I don't care 
> about ECOFF either.
> 

I can agree with that.

>  But this is not the right mailing list for people that actually use that 
> stuff, so it may be quite a good idea to ask at a more appropriate forum, 
> like <linux-mips@linux-mips.org>.

I've asked the netbsd people already, if you (or Thiemo) would like to
ask the linux people that'd be great.

I think we can deprecate the assembler support, but keep the bfd support
around. Likely, though, it'll bit rot quite a bit over the years unless
someone uses it.

-eric