This is the mail archive of the binutils@sourceware.org mailing list for the binutils project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: arm objdump bug?


Daniel Jacobowitz wrote:
On Fri, Sep 25, 2009 at 08:55:59AM -0500, Joel Sherrill wrote:
Hi,

We are doing object coverage analysis on RTEMS.  We
have been spending a lot of time looking at objdumps.
Our program which looks for nops and marks instructions
as not executed got confused when it saw this:


_Context_Switch( &_Thread_BSP_context, &_Thread_Heir->Registers );
f384: ea0015cc b 14abc <_CPU_Context_switch>
f388: 0001fd30 .word 0x0001fd30
f38c: 0001fa84 .word 0x0001fa84
f390: 0001fbd4 .word 0x0001fbd4
f394: 0001f7d8 .word 0x0001f7d8
f398: fbb4 .short 0xfbb4
f39a: 01 .byte 0x01
...

Typically, this means that the byte at f39b isn't just zero - it's unreadable for some reason, possibly because it isn't part of a section, or it's marked code instead of data (bad mapping symbols). There've been recent fixes for the mapping symbol code too.


OK. Weird. f39c is the start of another method in the same
section.

Unfortunately the machine producing these is using binutils 2.19.1
RTEMS RPMs. I will update to the RPMs based upon 2.19.90
and see if this issue and the i386 turn up again in another
round of testing.

Hopefully the latest binutils doesn't have this or the i386 quirk.

--joel


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]