Link RAP file error on zynq BSP

[jameszxj]

On 12/10/18 10:05 am, Joel Sherrill wrote:
>> 
>> 
>> On Thu, Oct 11, 2018 at 11:45 PM jameszxj <jameszxj at gmail.com
>> <mailto:jameszxj at gmail.com>> wrote:
>> 
>>     HI,
>>         I run RTEMS on my xilinx z7k board and dynamic load application(RAP
>>     format) from shell.
>>     In recent update, i encount a link error when link the RAP file.
>>     Here is the command and output message:
>>                     rtems-ld -n -e PT_Init --base rtems-nsa.elf
>>     wsp/build/nssys.elf -o wsp/build/nssys.out
>>                     error: rap::object: Section index '65522' not found:
>>     wsp/build/nssys.elf
>> 
>>         At last I found the bsp_spec file has been modified,  "*link:%{!qrtems:
>>     %(old_link)} %{qrtems: -dc -dp -N}" was removed
>>     from the bsp_spec file.  Add this line to the file, ld is OK.
>>        Why removed these lines?
>> 
>> 
>> The long-term goal is to remove the bsp_specs. I made a sweep making them more
>> consistent and have a set of patches against gcc master which eliminates them on
>> many architectures. Working so the details are implied by GCC.
>> 
>> So let's look at the options and see which one has an impact which mattered that
>> I didn't catch. This is the description of these options from ld's man page:
>> 
>>       -N
>>        --omagic
>>            Set the text and data sections to be readable and writable.  Also,
>>            do not page-align the data segment, and disable linking against
>>            shared libraries.  If the output format supports Unix style magic
>>            numbers, mark the output as "OMAGIC". Note: Although a writable
>>            text section is allowed for PE-COFF targets, it does not conform to
>>            the format specification published by Microsoft.
>> 
>>       -d
>>       -dc
>>        -dp These three options are equivalent; multiple forms are supported
>>            for compatibility with other linkers.  They assign space to common
>>            symbols even if a relocatable output file is specified (with -r).
>>            The script command "FORCE_COMMON_ALLOCATION" has the same effect.
>> 
>> I'm going to guess that it is -dc/-dp but have no idea how this would impact RAP
>> except to confuse the code. 
>> 
>> Can you try adding adding the options one at a time until you make it work? Then
>> Chris can figure out why it breaks RAP.>


I have tried this,just add -dc/-dp do not make it work.


>I suspect there is a section that is present which is needed that is not being
>handled in the RAP toolkit code. That RAP format merges some sections in rtems-ld.>


Probably like this, and I just test it use just one simple source file, uninitialized variable
cause this error. If I initialize the variable, link is fine.


>James, if you can make a test file that generates the error then create a ticket
>and attach the code I can take a look.>


>> 
>>        Does the newest RTEMS support the loading of incrementally linked ELF
>>     files? RAP format runs OK, but the link needs BSP image, this
>>     make BSP's development and APP's development not so independent. a little
>>     pity, :)
>> 
>> This is a Chris question.
>> 
>James, if you do not provide a base image, ie no --base option, what happens?>


I am trying to figure out the problem, it maybe has sth. to do with the --nostdlib option,
I need to do more tests. I shall describe the problem more clear later on.


>Chris
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.rtems.org/pipermail/devel/attachments/20181015/0b5c99ac/attachment-0002.html>