New ARC linker testsuite failure

New ARC linker testsuite failure

[Nick Clifton]

Hi Andrew,

  I am seeing a new failure in the linker testsuite for an arc-elf
  toolchain running on an x86_64 linux host:

regexp_diff match failure
regexp "^ *[0-9a-f]+:	5948 3000           	xldb	r10,\[0x57f03000\]$"
line   " 100:	3000 0000           	fmul	r0,r0,r0"
regexp_diff match failure
regexp "^ *[0-9a-f]+:	5949 3000           	xldw	r10,\[0x57f03000\]$"
line   " 104:	3000 0000           	fmul	r0,r0,r0"
regexp_diff match failure
regexp "^ *[0-9a-f]+:	594a 3000           	xld	r10,\[0x57f03000\]$"
line   " 108:	3000 0000           	fmul	r0,r0,r0"
regexp_diff match failure
regexp "^ *[0-9a-f]+:	594c 3000           	xstb	r10,\[0x57f03000\]$"
line   " 10c:	3000 0000           	fmul	r0,r0,r0"
regexp_diff match failure
regexp "^ *[0-9a-f]+:	594d 3000           	xstw	r10,\[0x57f03000\]$"
line   " 110:	3000 0000           	fmul	r0,r0,r0"
regexp_diff match failure
regexp "^ *[0-9a-f]+:	594e 3000           	xst	r10,\[0x57f03000\]$"
line   " 114:	3000 0000           	fmul	r0,r0,r0"
FAIL: ld-arc/nps-1a

  Can you reproduce this ?  If so, please could you fix it.

Cheers
  Nick

RE: New ARC linker testsuite failure

[Claudiu Zissulescu]

Hi Nick,

Issue confirmed. The fmul opcode is overlapping with nps400' x* opcodes. Andrew needs to make a better filter in the disassembler for those cases. 

Andrew please confirm if you will fix this issue, or I need to handle it.

Thanks,
Claudiu

> -----Original Message-----
> From: Nick Clifton [mailto:nickc@redhat.com]
> Sent: Friday, April 15, 2016 10:18 AM
> To: andrew.burgess@embecosm.com
> Cc: binutils@sourceware.org; Claudiu.Zissulescu@synopsys.com
> Subject: New ARC linker testsuite failure
> 
> Hi Andrew,
> 
>   I am seeing a new failure in the linker testsuite for an arc-elf
>   toolchain running on an x86_64 linux host:
> 
> regexp_diff match failure
> regexp "^ *[0-9a-f]+:	5948 3000           	xldb	r10,\[0x57f03000\]$"
> line   " 100:	3000 0000           	fmul	r0,r0,r0"
> regexp_diff match failure
> regexp "^ *[0-9a-f]+:	5949 3000           	xldw	r10,\[0x57f03000\]$"
> line   " 104:	3000 0000           	fmul	r0,r0,r0"
> regexp_diff match failure
> regexp "^ *[0-9a-f]+:	594a 3000           	xld	r10,\[0x57f03000\]$"
> line   " 108:	3000 0000           	fmul	r0,r0,r0"
> regexp_diff match failure
> regexp "^ *[0-9a-f]+:	594c 3000           	xstb	r10,\[0x57f03000\]$"
> line   " 10c:	3000 0000           	fmul	r0,r0,r0"
> regexp_diff match failure
> regexp "^ *[0-9a-f]+:	594d 3000           	xstw	r10,\[0x57f03000\]$"
> line   " 110:	3000 0000           	fmul	r0,r0,r0"
> regexp_diff match failure
> regexp "^ *[0-9a-f]+:	594e 3000           	xst	r10,\[0x57f03000\]$"
> line   " 114:	3000 0000           	fmul	r0,r0,r0"
> FAIL: ld-arc/nps-1a
> 
>   Can you reproduce this ?  If so, please could you fix it.
> 
> Cheers
>   Nick

Re: New ARC linker testsuite failure

[Andrew Burgess]

* Claudiu Zissulescu <Claudiu.Zissulescu@synopsys.com> [2016-04-15 09:11:53 +0000]:

> Hi Nick,
> 
> Issue confirmed. The fmul opcode is overlapping with nps400' x* opcodes. Andrew needs to make a better filter in the disassembler for those cases. 
> 
> Andrew please confirm if you will fix this issue, or I need to
> handle it.

I will investigate this issue this week.

Thanks for bringing this to my attention.

Andrew



> 
> Thanks,
> Claudiu
> 
> > -----Original Message-----
> > From: Nick Clifton [mailto:nickc@redhat.com]
> > Sent: Friday, April 15, 2016 10:18 AM
> > To: andrew.burgess@embecosm.com
> > Cc: binutils@sourceware.org; Claudiu.Zissulescu@synopsys.com
> > Subject: New ARC linker testsuite failure
> > 
> > Hi Andrew,
> > 
> >   I am seeing a new failure in the linker testsuite for an arc-elf
> >   toolchain running on an x86_64 linux host:
> > 
> > regexp_diff match failure
> > regexp "^ *[0-9a-f]+:	5948 3000           	xldb	r10,\[0x57f03000\]$"
> > line   " 100:	3000 0000           	fmul	r0,r0,r0"
> > regexp_diff match failure
> > regexp "^ *[0-9a-f]+:	5949 3000           	xldw	r10,\[0x57f03000\]$"
> > line   " 104:	3000 0000           	fmul	r0,r0,r0"
> > regexp_diff match failure
> > regexp "^ *[0-9a-f]+:	594a 3000           	xld	r10,\[0x57f03000\]$"
> > line   " 108:	3000 0000           	fmul	r0,r0,r0"
> > regexp_diff match failure
> > regexp "^ *[0-9a-f]+:	594c 3000           	xstb	r10,\[0x57f03000\]$"
> > line   " 10c:	3000 0000           	fmul	r0,r0,r0"
> > regexp_diff match failure
> > regexp "^ *[0-9a-f]+:	594d 3000           	xstw	r10,\[0x57f03000\]$"
> > line   " 110:	3000 0000           	fmul	r0,r0,r0"
> > regexp_diff match failure
> > regexp "^ *[0-9a-f]+:	594e 3000           	xst	r10,\[0x57f03000\]$"
> > line   " 114:	3000 0000           	fmul	r0,r0,r0"
> > FAIL: ld-arc/nps-1a
> > 
> >   Can you reproduce this ?  If so, please could you fix it.
> > 
> > Cheers
> >   Nick

Re: New ARC linker testsuite failure

[Andrew Burgess]

* Claudiu Zissulescu <Claudiu.Zissulescu@synopsys.com> [2016-04-15 09:11:53 +0000]:

> Issue confirmed. The fmul opcode is overlapping with nps400' x*
> opcodes. Andrew needs to make a better filter in the disassembler
> for those cases.
> 
> Andrew please confirm if you will fix this issue, or I need to
> handle it.

I believe the patch below is all that is required to resolve this
issue.  OK to apply?

Thanks,
Andrew

---

arc: Fix relocation formula for ARC_NPS_CMEM16 relocation

The ME modifier was missing from the relocation formula for the
ARC_NPS_CMEM16 relocation, and as such the relocation would not patch
correctly on little endian targets.

include/ChangeLog:

	* elf/arc-reloc.def (ARC_NPS_CMEM16): Add ME modifier to formula.
---
 include/ChangeLog         | 4 ++++
 include/elf/arc-reloc.def | 2 +-
 2 files changed, 5 insertions(+), 1 deletion(-)

diff --git a/include/elf/arc-reloc.def b/include/elf/arc-reloc.def
index 10703d2..faa1389 100644
--- a/include/elf/arc-reloc.def
+++ b/include/elf/arc-reloc.def
@@ -496,4 +496,4 @@ ARC_RELOC_HOWTO(ARC_NPS_CMEM16, 78, \
                 16, \
                 replace_bits16, \
                 dont, \
-                ( S + A ))
+                ( ME ( S + A )))
-- 
2.5.1

RE: New ARC linker testsuite failure

[Claudiu Zissulescu]

Hi Andrew,

> include/ChangeLog:
> 
> 	* elf/arc-reloc.def (ARC_NPS_CMEM16): Add ME modifier to
> formula.

It is alright. Do you have a test which flexes a  bit this newly introduced relocation also in GAS?

Best,
Claudiu

Re: New ARC linker testsuite failure

[Andrew Burgess]

* Claudiu Zissulescu <Claudiu.Zissulescu@synopsys.com> [2016-04-19 16:57:03 +0000]:

> > include/ChangeLog:
> > 
> > 	* elf/arc-reloc.def (ARC_NPS_CMEM16): Add ME modifier to
> > formula.
> 
> It is alright. Do you have a test which flexes a bit this newly
> introduced relocation also in GAS?

Yes, nps400-3.{d,s} is the assembler test.  Is that what you had in
mind?

Thanks,
Andrew

RE: New ARC linker testsuite failure

[Claudiu Zissulescu]

> Yes, nps400-3.{d,s} is the assembler test.  Is that what you had in
> mind?

If  this reloc will never be fixed by gas then it is enough.

Best,
Claudiu

Re: New ARC linker testsuite failure

[Nick Clifton]

Hi Andrew,

> include/ChangeLog:
> 
> 	* elf/arc-reloc.def (ARC_NPS_CMEM16): Add ME modifier to formula.

Approved - please apply.

Cheers
  Nick

RE: New ARC linker testsuite failure

[Claudiu Zissulescu]

> Thanks - please apply the patch.
> 

Committed. Thank you,
Claudiu

Re: New ARC linker testsuite failure

[Nick Clifton]

Hi Claudiu,

> I've fixed the indicated error plus additional two big-endian ARC errors. Please find the patch attached.

Thanks - please apply the patch.

Cheers
  Nick

Re: New ARC linker testsuite failure

[Claudiu Zissulescu]

[-- Attachment #1: Type: text/plain, Size: 137 bytes --]

Hi Nick,

I've fixed the indicated error plus additional two big-endian ARC 
errors. Please find the patch attached.

Thank you,
Claudiu

[-- Attachment #2: 0001-ARC-Fix-big-endian-ld-tests-errors.patch --]
[-- Type: text/plain, Size: 1867 bytes --]

From 7bc6f91e3c38c5cd7612a14696f509058594ce24 Mon Sep 17 00:00:00 2001
From: Claudiu Zissulescu <claziss@gmail.com>
Date: Thu, 20 Jul 2017 12:47:30 -0400
Subject: [PATCH] [ARC] Fix big-endian ld tests errors.

A set of small patches that are fixing big-endian observed errors.

ld/
2017-07-20  Claudiu Zissulescu  <claziss@synopsys.com>

	* testsuite/ld-arc/jli-overflow.d: Force testing for little
	endian.
	* testsuite/ld-arc/tls_gd-01.d: Fix string to match bigendian
	systems.
	* testsuite/ld-arc/tls_ie-01.d: Fix test for bigendian systems.
---
 ld/testsuite/ld-arc/jli-overflow.d | 4 ++--
 ld/testsuite/ld-arc/tls_gd-01.d    | 2 +-
 ld/testsuite/ld-arc/tls_ie-01.d    | 4 ++--
 3 files changed, 5 insertions(+), 5 deletions(-)

diff --git a/ld/testsuite/ld-arc/jli-overflow.d b/ld/testsuite/ld-arc/jli-overflow.d
index b1a2663..6445b2a 100644
--- a/ld/testsuite/ld-arc/jli-overflow.d
+++ b/ld/testsuite/ld-arc/jli-overflow.d
@@ -1,4 +1,4 @@
 #source: jli-overflow.s
-#as:
-#ld: -q -T jli-script.ld
+#as: -EL
+#ld: -q -T jli-script.ld -A elf32-arclittle
 #error_output: jli-overflow.err
diff --git a/ld/testsuite/ld-arc/tls_gd-01.d b/ld/testsuite/ld-arc/tls_gd-01.d
index e789157..d65f2cd 100644
--- a/ld/testsuite/ld-arc/tls_gd-01.d
+++ b/ld/testsuite/ld-arc/tls_gd-01.d
@@ -4,7 +4,7 @@
 #objdump: -d
 #xfail: arc*-*-elf*
 
-[^:]+:     file format elf32-littlearc
+[^:]+:     file format elf32-.*arc
 
 
 Disassembly of section \.text:
diff --git a/ld/testsuite/ld-arc/tls_ie-01.d b/ld/testsuite/ld-arc/tls_ie-01.d
index 62b3932..e4af7c2 100644
--- a/ld/testsuite/ld-arc/tls_ie-01.d
+++ b/ld/testsuite/ld-arc/tls_ie-01.d
@@ -4,7 +4,7 @@
 #objdump: -s -j .got
 #xfail: arc*-*-elf*
 
-[^:]+:     file format elf32-littlearc
+[^:]+:     file format elf32-.*arc
 
 Contents of section \.got:
- [0-9a-f]+ 08000000 0c000000 +.+
+ [0-9a-f]+ [08]+ [0c]+ +.+
-- 
2.9.3

RE: New ARC linker testsuite failure

[Claudiu Zissulescu]

Sure, sorry for this. 

Claudiu

> -----Original Message-----
> From: Nick Clifton [mailto:nickc@redhat.com]
> Sent: Wednesday, July 19, 2017 3:49 PM
> To: Claudiu.Zissulescu@synopsys.com; John.Martin@emmicro-us.com
> Cc: binutils@sourceware.org
> Subject: New ARC linker testsuite failure
> 
> Hi Claudiu, Hi John,
> 
>   A new linker testsuite failure has just shown up:
> 
> FAIL: ld-arc/jli-overflow
> 
>   This is only with the arceb-elf target and the log file appears to
>   suggest that you need to add an endian specifier when linking:
> 
> ./ld-new: tmpdir/jli-overflow.o: compiled for a big endian system and target
> is little endian
> ./ld-new: failed to merge target specific data of file tmpdir/jli-overflow.o
> tmpdir/jli-overflow.o: In function `test_1025':
> (.text+0x2004): relocation truncated to fit: R_ARC_JLI_SECTOFF against
> symbol `__jli.func1025' defined in .jlitab[.jlitab.func1025] section in tmpdir/jli-
> overflow.o
> ./ld-new: final link failed: Symbol needs debug section which does not exist
> 
>   Could you take a look at this please ?
> 
> Cheers
>   Nick

New ARC linker testsuite failure

[Nick Clifton]

Hi Claudiu, Hi John,

  A new linker testsuite failure has just shown up:

FAIL: ld-arc/jli-overflow

  This is only with the arceb-elf target and the log file appears to
  suggest that you need to add an endian specifier when linking:

./ld-new: tmpdir/jli-overflow.o: compiled for a big endian system and target is little endian
./ld-new: failed to merge target specific data of file tmpdir/jli-overflow.o
tmpdir/jli-overflow.o: In function `test_1025':
(.text+0x2004): relocation truncated to fit: R_ARC_JLI_SECTOFF against symbol `__jli.func1025' defined in .jlitab[.jlitab.func1025] section in tmpdir/jli-overflow.o
./ld-new: final link failed: Symbol needs debug section which does not exist

  Could you take a look at this please ?

Cheers
  Nick