Release 2.22: Next week ...

[Matthew Gretton-Dann]

On 17/11/11 15:19, Tristan Gingold wrote:
>
> On Nov 17, 2011, at 4:10 PM, Matthew Gretton-Dann wrote:
>
>> On 17/11/11 14:50, Tristan Gingold wrote:
>>> On Nov 9, 2011, at 10:25 AM, Tristan Gingold wrote:
>>>
>>>> Hi,
>>>>
>>>> I plan to make the 2.22 release next week.
>>>> There are still a few pending back ports: micromips and 'V2 of get symbol linker plugin interface' come to my mind.
>>>>
>>>> As usual, speak up if you think I missed something.
>>>
>>> I have run the testsuite for most of the targets.  I think there aren't blocking regressions for a 22.0, although I think it would be nice to clean-up arm-eabi and mips*-* for the 22.1
>>
>>> arm-eabi: FAIL: R_ARM_THM_JUMP24 Relocation veneers: Long
>>> arm-eabi: FAIL: R_ARM_THM_JUMP24 Relocation veneers: Short 1
>>> arm-eabi: FAIL: R_ARM_THM_JUMP24 Relocation veneers: Short 2
>>
>> I thought I had fixed these with:
>> 	http://sourceware.org/ml/binutils/2011-11/msg00072.html
>>
>> Which was committed:
>> 	http://sourceware.org/ml/binutils-cvs/2011-11/msg00042.html
>>
>> When I run the tests myself targetting arm-none-eabi they pass, as they also (I am assuming) in your arm-linux-gnueabi run (you didn't highlight them there).
>
> Yes, I was surprised too.
>
>> Do you mind sending me the appropriate part of ld.log from the testrun and I will try to see if I can reproduce the failure?
>
> Sure:
>
> /tmp/binutils-2.21.90-sc/objs-arm-eabi/ld/../gas/as-new    -o tmpdir/jump-reloc-veneers.o -march=armv7-a -mthumb /tmp/binutils-2.21.90-sc/binutils-2.21.90/ld/testsuite/ld-arm/jump-reloc-veneers.s
> Executing on host: sh -c {/tmp/binutils-2.21.90-sc/objs-arm-eabi/ld/../gas/as-new    -o tmpdir/jump-reloc-veneers.o -march=armv7-a -mthumb /tmp/binutils-2.21.90-sc/binutils-2.21.90/ld/testsuite/ld-arm/jump-reloc-veneers.s 2>&1}  /dev/null ld.tmp (timeout = 300)
> /tmp/binutils-2.21.90-sc/objs-arm-eabi/ld/ld-new   -o tmpdir/jump-reloc-veneers-short1 -L/tmp/binutils-2.21.90-sc/binutils-2.21.90/ld/testsuite/ld-arm --no-fix-arm1176 --section-start destsect=0x00009000 --section-start .text=0x8000 tmpdir/jump-reloc-veneers.o
> Executing on host: sh -c {/tmp/binutils-2.21.90-sc/objs-arm-eabi/ld/ld-new   -o tmpdir/jump-reloc-veneers-short1 -L/tmp/binutils-2.21.90-sc/binutils-2.21.90/ld/testsuite/ld-arm --no-fix-arm1176 --section-start destsect=0x00009000 --section-start .text=0x8000 tmpdir/jump-reloc-veneers.o 2>&1}  /dev/null ld.tmp (timeout = 300)
> Executing on host: sh -c {/tmp/binutils-2.21.90-sc/objs-arm-eabi/ld/../binutils/objdump -d tmpdir/jump-reloc-veneers-short1>dump.out 2>ld.stderr}  /dev/null  (timeout = 300)
> /tmp/binutils-2.21.90-sc/objs-arm-eabi/ld/../binutils/objdump -d tmpdir/jump-reloc-veneers-short1
> regexp_diff match failure
> regexp "^    9000:      e7fe            b.n     9000<dest>$"
> line   "    9000:       f7ff bffe       b.w     9000<dest>"
> FAIL: R_ARM_THM_JUMP24 Relocation veneers: Short 1

So this looks like the failure mode I fixed.

> Tell me if you need more.

I still can't reproduce this - can you send me you configure 
command-line and the host you are running on please?

I don't think this is a reason to hold the release up - the failure is 
not a failure in the linker, but rather the test not keeping up with a 
(correct) change in the assembler's output.

Thanks,

Matt


-- 
Matthew Gretton-Dann
Principal Engineer, PD Software - Tools, ARM Ltd