sptls01 fails on sis

[Joel Sherrill]

This fails on sis. Marking for milestone 4.11 until Sebastian gets a chance to look at it and decide if this should work.

Starting program: /users/joel/rtems-4.11-work/b-sis/sparc-rtems4.11/c/sis/testsuites/sptests/sptls01/sptls01.exe

* BEGIN OF TEST SPTLS 1 *
TLS item = 0
../../../../../../../rtems/c/src/../../testsuites/sptests/sptls01/init.c: 36 tls_item == expected
[Inferior 1 (process 42000) exited normally]

[Sebastian Huber]

I think this is a problem with the GDB patches. I don't see this problem with a vanilla GDB 7.8.2.

[Jiri Gaisler]

This is a problem in the updated sis, which keeps the emulated memory in host endian. If a elf section's length is not a multiple of 4, the last data does not get correct endian when loaded into the simulator. The sptls01 has a section of 1 byte, hence the problem:

section: .text, addr: 0x2000000, size 96304 bytes
section: .tdata, addr: 0x2017830, size 1 bytes <====!
section: .data, addr: 0x2017840, size 1632 bytes

I will make a patch for this, but I do not know when it will be merged in gdb git repo ...

Jiri.

[Sebastian Huber]

Is this problem fixed?

[Joel Sherrill]

I think Jiri fixed it but the patch set is not completely merged in gdb. I should check that all SPARC bsps pass with the rsb version.

Let's leave this open until at least I conforim the rsb.

[Jiri Gaisler]

This problem is fixed and merged in gdb git. Should be available in gdb-7.9.1. Outstanding patches are use-gdb-callback-for-UART-IO and support for leon2/3.

[Sebastian Huber]

Seems to be fixed.