ARM testsuite failures...

ARM testsuite failures...

[Scott Bambrough]

/home/scottb/cvstree/binutils-build/ld/ld-new: error: cannot change
output format whilst linking ARM binaries
FAIL: S-records

/home/scottb/cvstree/binutils-build/ld/ld-new: error: cannot change
output format whilst linking ARM binaries
FAIL: S-records with constructors

It would seem these tests are expected to fail on ARM platforms.  I have
attached a patch that sets them up as expected failures for ARM Linux as
well.

Scott

1999-10-28  Scott Bambrough  <scottb@netwinder.org>

        * ld/testsuite/ld-srec/srec.exp: Setup expected failures for
	  srec tests on ARM Linux.
Index: srec.exp
===================================================================
RCS file: /cvs/binutils/binutils/ld/testsuite/ld-srec/srec.exp,v
retrieving revision 1.2
diff -u -r1.2 srec.exp
--- srec.exp	1999/05/11 20:31:53	1.2
+++ srec.exp	1999/10/28 15:56:51
@@ -229,7 +229,7 @@
     }
 
     # ARM targets call __gccmain
-    if [istarget arm-*-*] {
+    if [istarget arm*-*-*] {
 	set flags "$flags --defsym ___gccmain=0"
     }
     
@@ -332,6 +332,7 @@
 setup_xfail "thumb-*-coff*"
 setup_xfail "thumb-*-pe*"
 setup_xfail "thumb-*-elf*"
+setup_xfail "arm*-*-linux*"
 
 # The S-record linker doesn't build special EABI sections.
 setup_xfail "powerpc*-*-eabi*"
@@ -368,7 +369,7 @@
 setup_xfail "i*86-*-aout*"
 setup_xfail "mips*-*-irix5*" "mips*-*-irix6*"
 setup_xfail "*-*-aix*" "*-*-xcoff*"
-setup_xfail "arm-*-*"
+setup_xfail "arm*-*-*"
 setup_xfail "strongarm-*-*"
 setup_xfail "thumb-*-*"
 setup_xfail "powerpc*-*-eabi*"

Re: ARM testsuite failures...

[Catherine Moore]

: However the failure of selective6 has been with use for some time.  I
: belive that Catherine had a fix for it though, so maybe she can
: comment.

   It's a bogus test.  I'm going to take it out of the testsuite a little
   later today.  You can ignore the failure, Scott.

   --catherine

Re: ARM testsuite failures...

[Nick Clifton]

Hi Scott,

: I cannot find any documentation on what --gc-sections is supposed to
: do.

From ld/ld.texinfo:

  @itemx --gc-sections
  Enable garbage collection of unused input sections.  It is ignored
  on targets that do not support this option.  This option is not
  compatible with @samp{-r}, nor should it be used with dynamic
  linking.  The default behaviour (of not performing this garbage
  collection) can be restored by specifying @samp{--no-gc-sections} on
  the command line.

: Should this test even be run for targets arm*-unknown-linux-gnu?

I don't see why not.  This is a test of the compiler and linker and it
is not OS specific.  (It is file format specific, though.  It only
works for ELF.  But arm-linux uses ELF doesn't it ? )

However the failure of selective6 has been with use for some time.  I
belive that Catherine had a fix for it though, so maybe she can
comment.

Cheers
	Nick

ARM testsuite failures...

[Scott Bambrough]

Hi all,

I have just one test suite failure to resolve on ARM Linux, and I'm in
need of some information.  The test that fails is selective6, from
ld/ld-selective/selective.exp.   It fails as follows:

gcc -w -O2 -ffunction-sections -fdata-sections -fvtable-gc
-fno-exceptions -fno-rtti
-B/home/scottb/cvstree/binutils-build/ld/tmpdir/gas/
-I/home/scottb/cvstree/binutils/ld/testsuite/ld-selective -g -O2 -W
-Wall  -c /home/scottb/cvstree/binutils/ld/testsuite/ld-selective/5.cc
-o tmpdir/5.o
/home/scottb/cvstree/binutils-build/ld/ld-new  -o tmpdir/5.x
--gc-sections -Bstatic tmpdir/5.o
/home/scottb/cvstree/binutils-build/ld/../binutils/nm-new  tmpdir/5.x
>tmpdir/nm.out
Setting nm_output(*ABS*) to 0x00000000
Setting nm_output(GLOBAL_.I.foo__1A) to 0x0200010c
Setting nm_output(bss_end__) to 0x02008160
Setting nm_output(bss_start) to 0x02008158
Setting nm_output(bss_start__) to 0x02008158
Setting nm_output(data_start) to 0x0200812c
Setting nm_output(end__) to 0x02008160
Setting nm_output(static_initialization_and_destruction_0) to 0x020000bc
Setting nm_output(vt_1A) to 0x02008140
Setting nm_output(vt_1B) to 0x0200812c
Setting nm_output(bss_end__) to 0x02008160
Setting nm_output(edata) to 0x02008158
Setting nm_output(end) to 0x02008160
Setting nm_output(etext) to 0x0200012c
Setting nm_output(start) to 0x0200009c
Setting nm_output(a) to 0x02008158
Setting nm_output(b) to 0x0200815c
Setting nm_output(foo__1A) to 0x02000074
Setting nm_output(getme__Fv) to 0x02000084
foo__1A== 0x02000074
FAIL: selective6

I cannot find any documentation on what --gc-sections is supposed to
do.  The only reference I can find to it is in an email from Catherine
Moore, and she only states that the reloc R_ARM_GNU_VTENTRY is used to
support the linker option --gc-sections.  I don't believe ARM Linux uses
this reloc.  Should this test even be run for targets
arm*-unknown-linux-gnu?  I'd appreciate it if someone could point me to
the documentation for this switch if I have missed it, or describe what
it is supposed to accomplish.

Thanks,

Scott

-- 
Scott Bambrough - Software Engineer
REBEL.COM    http://www.rebel.com
NetWinder    http://www.netwinder.org

Re: ARM testsuite failures...

[Nick Clifton]

Hi Scott,

: 1999-10-28  Scott Bambrough  <scottb@netwinder.org>
: 
:         * ld/testsuite/ld-srec/srec.exp: Setup expected failures for
: 	  srec tests on ARM Linux.

Accpeted and applied.

Thanks,

	Nick