Re: MIPS simulator is broken

[Steve Ellcey <sellcey@imgtec.com>]

On Fri, 2016-01-08 at 17:48 -0500, Mike Frysinger wrote:
> On 08 Jan 2016 14:30, Steve Ellcey  wrote:
> > Your December 26th change to the mips simulator has broken it, at least
> > for me.  I am trying to compile and run a program with mips32r2 big-endian
> > and am using the mti32.ld linker script.
> > 
> > After your commit my attempts to run a 'hello world' program die with:
> 
> please add a testcase to the sim/testsuite/ tree that triggers the error
> when you run `make check`.  i can look after that.
> -mike

I have no idea how to add a testcase to sim/testsuite but the mips
tests that are already there all fail (they used to work).  I built the
mips-sde-elf toolchain (to see if it worked better than mips-mti-elf)
and when I ran 'make check' in the sim directory all the existing mips
tests failed on execution.  I retested after backing out your change and
everything passed.

An example of one failure:

/scratch/sellcey/repos/elf/obj-mips-sde-elf/gdb/sim/testsuite/../mips/run     sanity.s.x
spawn /scratch/sellcey/repos/elf/obj-mips-sde-elf/gdb/sim/testsuite/../mips/run sanity.s.x
mips-core: 4 byte read to unmapped address 0xffffffff80010000 at 0xffffffff80010000
program stopped with signal 10 (User defined signal 1).
FAIL: mips32 sanity.s (execution)

Steve Ellcey
sellcey@imgtec.com