[LLVMdev] Debug info failing in assembler.

Richard Pennington rich at pennware.com
Sat Sep 21 20:27:55 PDT 2013


I filed PR 17321 for this.
The problem code is in llvm/lib/CodeGen/AsmPrinter/DwarfDebug.cpp around 
line 2844.  Apparently Span.End is set for the .comm symbol in this case.

-Rich

On 09/21/2013 10:26 PM, Richard Mitton wrote:
> If it thinks the symbol is in the BSS section, then it should never 
> have tried to use .comm to emit it I think.
>
> On x86 it does not try to mix and match, which is why it works. AFAIK 
> comm symbols are regarded as having no section, rather than being bss, 
> so I think it's a bug in whatever code printed that .comm statement.
>
> I'll look into this tomorrow.
>
>> Eric Christopher <mailto:echristo at gmail.com>
>> Saturday, September 21, 2013 11:26 AM
>>
>> Interesting. File please? Thanks.
>>
>> _______________________________________________
>> LLVM Developers mailing list
>> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev
>> Richard Pennington <mailto:rich at pennware.com>
>> Saturday, September 21, 2013 5:52 AM
>> Hi,
>>
>> I just updated from r190763 to r191137 and started getting failures 
>> in generated assembly language when debug info is enabled. Here is 
>> the test case:
>>
>> // Compile and run for every target.
>> // RUN: %ecc -g -o %t %s && %t
>> // FAIL: %armecc -g -o %t %s && %armrun %t
>> // FAIL: %armebecc -g -o %t %s && %armebrun %t
>> // RUN: %i386ecc -g -o %t %s && %i386run %t
>> // FAIL: %microblazeecc -g -o %t %s && %microblazerun %t
>> // FAIL: %mipsecc -g -o %t %s && %mipsrun %t
>> // FAIL: %mipselecc -g -o %t %s && %mipselrun %t
>> // FAIL: %ppcecc -g -o %t %s && %ppcrun %t
>> // FAIL: %ppc64ecc -g -o %t %s && %ppc64run %t
>> // RUN: %x86_64ecc -g -o %t %s && %x86_64run %t
>>
>> int comm;
>> int main()
>> {
>>     comm = 6;
>> }
>>
>> This fails on all but the x86 processors. The typical failure looks 
>> like this:
>> ...
>>         .type   comm, at object            # @comm
>>         .comm   comm,4,4
>>         .cfi_sections .debug_frame
>>         .bss
>> .L.bss_end:
>> ...
>> .L.debug_abbrev_end:
>>         .section        .debug_aranges,"", at progbits
>>         .long   36                      # Length of ARange Set
>>         .short  2                       # DWARF Arange version number
>>         .long   .L.debug_info_begin0    # Offset Into Debug Info Section
>>         .byte   4                       # Address Size (in bytes)
>>         .byte   0                       # Segment Size (in bytes)
>>         .byte   255
>>         .byte   255
>>         .byte   255
>>         .byte   255
>>         .long   comm
>> .Lset0 = .L.bss_end-comm
>>         .long   .Lset0
>>
>> The typical error message is:
>> /tmp/debug-8ecf9d.s: Assembler messages:
>> /tmp/debug-8ecf9d.s: Error: invalid operands (.bss and *COM* 
>> sections) for `-' when setting `.Lset0'
>>
>> Is it legal to mix .comm symbols and .bss symbols in assembly? Is the 
>> x86 success a bug or a feature?
>>
>> -Rich
>> _______________________________________________
>> LLVM Developers mailing list
>> LLVMdev at cs.uiuc.edu http://llvm.cs.uiuc.edu
>> http://lists.cs.uiuc.edu/mailman/listinfo/llvmdev

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20130921/a3b414ea/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: compose-unknown-contact.jpg
Type: image/jpeg
Size: 770 bytes
Desc: not available
URL: <http://lists.llvm.org/pipermail/llvm-dev/attachments/20130921/a3b414ea/attachment.jpg>


More information about the llvm-dev mailing list