Cherrypick from master 2003-08-13 17:48:27 UTC Joel Sherrill <joel.sherrill@OARcorp.com> '2003-08-13 Joel Sherrill <joel@OARcorp.com>':
c/src/tests/samples/fileio/.cvsignore
PR 369/filesystem
* Makefile.am, configure.ac: sample application to show the use of the
DOSFS functions
* fileio/Makefile.am, fileio/fileio.doc, fileio/init.c,
fileio/system.h: New files.
PR 437/bsps
* irq/irc.c: calls to bspIo/printk must not use '%i' format which is
apparently not supported
* pci/pci.c: calls to bspIo/printk must not use '%i' format which is
apparently not supported
PR 288/rtems
* irq/irq_asm.S, startup/bspstart.c: _ISR_Nest_level is now properly
maintained and any BSP which does not do so will trip a panic in the
shared code.
PR 415/bsps
* pci/detect_raven_bridge.c, startup/bspstart.c:
This patch reverts enabling MCP interrupts for the motorola_powerpc
BSPs. REASON: pci config space scanning (as e.g. done by the DEC
Ethernet driver) would raise machine check exceptions.
PR 429/bsps
PR 432/bsps
* network/elnk.c: Due to a bug in the driver, if multiple packets
are queued onto an elnk board, the ISR will deadlock the device with
a tx complete interrupt. Lighter tx loads are not affected as they
end up submitting single packets to the device.
PR 428/bsps
PR 432/bsps
* bootloader/pci.c: Re-instated code that prevents remapping small
IO regions, which if remapped would cause i8259 registers to move
out from under the #define'd base addresses.
* startup/bspstart.c: Reduced BAT2 PCI memory allocation to 256
megs, I incorrectly had extended it which would cause problems with
PCI devices that defined prefetchable memory.
PR 433/tools
* mips-stub.c: Patch from C. M. Heard <heard@pobox.com> described as:
The purpose of the patch is to make the stub ignore all
characters except '+' (ack) or '-' (nak) when it has transmitted a
packet and expects an acknowledgment. Without the patch,
putpacket() will retransmit each time it receives anything other
than '+' when expecting an ack, and that, bitter experience has
found, frequently results in the stub failing to sync up when
initially attaching to gdb.
I think that rtems users will notice significant improvement in
the of the mips stub if this patch is incorporated in the
distribution; certainly I did in my stand-aline environment. I
should hasten to add that it is only an improvement and not a 100%
cure for synchronization problems because '+' characters can (and
do) during as message data, and it's still possible for these to
be mistaken as acks. I have no idea how to fix that short of
redesigning the gdb-remote protocol.
