Custom Query (105 matches)

RSB fatal error on missing hash checksums (cloned)

Cloned from #4760:

The time has come to raise a fatal error if a hash is missing rather than a warning. This will stop missing checksums being present when releasing.

It also makes it simpler to test releases for no missing checksums.

Net-Snmp referenced patch missing

The Net-SNMP configuration references a patch that is missing. I cannot find it.

Update to 5.9.3 and for the libbsd stack.

RSB freetype 2.4.10 source not hosted any longer

The source is not present any longer. I think the simplest solution maybe using the source in the 5.1 release and pointing the RSB 5 freetype URL to that file.

Any updates would be a change in the code and functionality.

Docs build system does not build singlehtml

The configure logic does not handle the inliner detection correctly.

Set RSB GDB source path to https

Backport ​https://git.rtems.org/rtems-source-builder/commit/?id=949cf500b4f4ab560a77c2ac60cd9ac0f47aa852

RTEMS docs do not build with a recent sphinx

The build does not configure:

$ ./waf distclean configure --prefix=/opt/src/rtems/test --pdf --singlehtml
'distclean' finished successfully (0.000s)
Setting top to                           : /opt/work/chris/rtems/docs/rtems-docs.git 
Setting out to                           : /opt/work/chris/rtems/docs/rtems-docs.git/build 
Checking for program 'git'               : /usr/local/bin/git 
Checking for program 'sphinx-build'      : /opt/work/chris/rtems/releasing/rng/bin/sphinx-build 
Checking for program 'aspell'            : /usr/local/bin/aspell 
Checking if Sphinx is at least 1.3       : yes (5.3) 
Checking Sphinx Options                  : none 
Checking Sphinx Nit-Pick mode            : no 
Checking for 'sphinx.ext.autodoc'        : found 
Checking for 'sphinx.ext.coverage'       : found 
Checking for 'sphinx.ext.doctest'        : found 
Checking for 'sphinx.ext.graphviz'       : found 
Checking for 'sphinx.ext.intersphinx'    : found 
Checking for 'sphinx.ext.mathjax'        : found 
Checking for 'sphinxcontrib.bibtex'      : not found (see README.txt) 
The configuration failed
(complete log in /opt/work/chris/rtems/docs/rtems-docs.git/build/config.log)

RSB decode exception stops build

Cloned from #4726:

Building in a Rocky VM on FB 12 with 5 I got:

Traceback (most recent call last):
  File "/usr/lib64/python3.9/threading.py", line 973, in _bootstrap_inner
    self.run()
  File "/usr/lib64/python3.9/threading.py", line 910, in run
    self._target(*self._args, **self._kwargs)
  File "/opt/work/chris/rtems/rsb/rtems-source-builder.git/source-builder/sb/execute.py", line 204, in _readthread
    data = data.decode(sys.stdout.encoding)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xe2 in position 4095: unexpected end of data

If the data is corrupted or broken things stop. Fix to attempt to keep going.

This issue also effect rtems-tools.

Set top in RSB version.py

Cloned from #4732:

Setting the top lets the code be used in deployment to get the version of the RSB being used. An RSB version lets a user track the exact tools being built and version numbers reported by the tools can match the packaged build, eg an rpm name and version info.

rtems-source-builder doesn't generate tar archives for all packages any more

Cloned from #4730:

Like discussed in the following mailing list thread, the rtems-source-builder doesn't generate tar archives for all packages anymore:

​https://lists.rtems.org/pipermail/devel/2022-September/073369.html

All RTEMS tools build sets 6/rtems-* work fine except for 6/rtems-microblaze. devel/qemu and devel/dtc do not generate a tar archive. If the patch

​https://git.rtems.org/rtems-source-builder/commit/?id=6205068c5a429e9ee1b471f4c8a3a119bb6757b2

is reverted, the tar files are generated just like before.

RSB decode exception stops build

Building in a Rocky VM on FB 12 with 5 I got:

Traceback (most recent call last):
  File "/usr/lib64/python3.9/threading.py", line 973, in _bootstrap_inner
    self.run()
  File "/usr/lib64/python3.9/threading.py", line 910, in run
    self._target(*self._args, **self._kwargs)
  File "/opt/work/chris/rtems/rsb/rtems-source-builder.git/source-builder/sb/execute.py", line 204, in _readthread
    data = data.decode(sys.stdout.encoding)
UnicodeDecodeError: 'utf-8' codec can't decode byte 0xe2 in position 4095: unexpected end of data

If the data is corrupted or broken things stop. Fix to attempt to keep going.

Git commit message format instrunctions (cloned)

The Eng docs ​Creating A Patch references the Wiki. The content should be in the manual.

Note lots of git related bits are in the Eng manual and still on the Wiki so which is correct?

Add RSB Deployment support

Provide deployment support for RTEMS 5.

The support will be a single architecture with multiple BSPs for that architecture.

This change does not effect any existing functionality.

Backport RSB fixes from Development

Backport RSB fixes from the development branch.

RSB get sources misses a number of files

Back port the changes from the development branch. The get sources tool is used during the release to collect the sources so they are hosted on ​https://ftp.rtems.org

The changes exposes a number of configuration file issues on the 5 branch.

RSB does not expand dir type macros correctly (cloned)

Cloned from #4710:

The expansion of the dir type macros is not done correctly. Macro replacement only appends the file part to the last element of a path list. For example:

  %{mydir}/abc.cfg

with a path of:

  %define mydir x:y:z

results in:

  x:y:z/abc.cfg

The config load code handles this but it is wrong. The expansion of dir type macro variables should handle path separators (:) so you get:

 x/abc.cfg:y/abc.cfg:z.abc.cfg

This has not been a problem in the RSB because the _configdir has only had two elements and the last one only ever had subdirectories.

Installed header break C++

Some headers generate C++ errors and so cannot be included.

Fix invalid RSB source URLs

Some packages now have invalid source URLs

Documentation List for Releases Completely Missing 5.x

See ​https://docs.rtems.org/releases.html and fix it. :)

Leaked BSP section flags in Makefile.inc breaks EPICS

RTEMS 5 added -ffunction-sections -fdata-sections to the target flags for BSPs in an architecture that supports code and data sections.

Some EPICS code such as the ​NTP have code the linker does not link into the final executable. The data sections means the done variable is not linked into the executable when a function in the module is referenced.

RTEMS 6 only exports the ABI flags and any internal build flags are not exported.

EPICS should also review this type of code because it excludes potential optimizations.

incorrect handling of "inactive_per_block" from "Objects_Information" structure

The inactive_per_block for block 1 is not handled correctly. For the first object in the second block, the inactive_per_block value is not decremented. As a consequence, the block might be deleted when the second to last object from block 1 is deleted. This is only valid when the application is configured with "CONFIGURE_UNLIMITED_OBJECTS".

The reason is the following:

in function "_Objects_Activate_unlimited" from "objectimpl.h"

block = _Objects_Get_index( the_object->id ) - OBJECTS_INDEX_MINIMUM;

if ( block > objects_per_block ) {

block /= objects_per_block;

information->inactive_per_block[ block ]--; information->inactive--;

}

in a block with 8 objects per block, when creating the 9th object, _Objects_Get_index( the_object->id ) returns 9. Therefore, "block" will be 8 (9 - 1). The next "if" will not be taken (8 > 8 ? No) and the inactive_per_block will not be decremented to 7 and remains at 8. This means that block 1 is a candidate for "_Objects_Shrink_information" function although object id 9 is still valid.

Steps to reproduce: -assuming 8 objects per allocation block -using semaphores as objects

1. create 17 "RTEMS_SIMPLE_BINARY_SEMAPHORE" semaphore objects id's 1 -> 17.
2. delete semaphores 11, 12, 13, 14, 15, 16, 17 (order is important)
3. lock semaphore 9
4. delete semaphore 10 -> at this point, block 1 is freed although semaphore 9 is still active
5. unlock semaphore 9 -> this can lead to errors. In my case due to the free the semaphore object was altered and changed from a "SEMAPHORE_VARIANT_SIMPLE_BINARY" to "SEMAPHORE_VARIANT_MUTEX_INHERIT_PRIORITY" resulting in a return value of "RTEMS_NOT_OWNER_OF_RESOURCE".

Simple fix: the condition for the "if" should be ">=" instead of ">".

sync() whould synchronize all file descriptors

According to POSIX

​https://pubs.opengroup.org/onlinepubs/9699919799/functions/sync.html

we have

The sync() function shall cause all information in memory that updates file systems to be scheduled for writing out to all file systems.

Currently, the RTEMS sync() implementation synchronizes only the file descriptors associated with a FILE object. This should be changed to call fsync() and fdatasync() for all file descriptors.

pfctl: Fix global state initialization

The last update introduced a static variable in get_query_socket() which was not initialized.

if_atsam: Fix checksum offload, add multicast and VLAN support

The if_atsam network interface driver has several issues. Do not disable the FCS if the checksum offload is disabled. Make sure the interface capabilities are enabled. Add multicast and VLAN support. Use the interface transmit method instead of the interface start approach. Correct a potential receive starvation error. Fix the interface up/down.

tcpdump: Fix dumping to file and reading from file

When dumping to file, the file needs to be closed at program exit. When reading from file, we do not need the loop monitor. Do not use signals, since they are not cleaned up at program exit.

Atomic store does not use the order parameter for C++ (cloned)

Cloned from #4566:

This could result in invalid inline code for the SMP ticket locks if used from C++.

Timecounter: Add NTP support (cloned)

Cloned from #2348:

The FreeBSD timecounter implementation supports the NTP. This support is currently disabled in RTEMS.

gdb 9.1 no longer builds on Cygwin (thread local error)

The Cygwin link error is:

cp-support.o: in function `gdb_demangle(char const*, int)':
/home/Christian/binutils-gdb/obj/gdb/../../gdb/cp-support.c:1619:(.text+0x6472): relocation truncated to fit: R_X86_64_PC32 against undefined symbol `TLS init function for thread_local_segv_handler'
/home/Christian/binutils-gdb/obj/gdb/../../gdb/cp-support.c:1619:(.text+0x648b): relocation truncated to fit: R_X86_64_PC32 against undefined symbol `TLS init function for thread_local_segv_handler'
collect2: error: ld returned 1 exit status

The issue appears to be that Cygwin changed their base address for executables and there are multiple approaches to fixing this on the Internet. None of these applied cleanly to a vanilla gdb 9.1. I adapted the patch presented at ​https://stackoverflow.com/questions/61984974/failed-to-build-avr-and-arm-gdb-9-1-under-cygwin-relocation-truncated-to-fit

This patch avoids using a thread-local extern variable, which causes link errors on some platforms, notably Cygwin. The poster through this was a better pattern even outside of working around linker bugs because it encapsulates direct access to the variable inside the class, instead of having a global extern variable.

It should be noted that gdb 10.1 builds cleanly on Cygwin but discussions leaned against bumping the gdb on the 5 branch. Hence we end up with a Cygwin specific patch for gdb 9.1 on Cygwin.

Re-add lost capability for custom stack allocator to allocate IDLE thread stacks

The move to statically allocate the stacks for the IDLE threads resulted in the loss of the capability for a custom stack allocator to be able to allocate the idle threads' stacks.

Map LibBSD bus space to the PCI base address for motorola_powerpc BSP

This provides the per BSP support for a split PCI address space in LibBSD for this the MVME2307 BSP

Make [out/in] le and be calls conditional

These calls clash with the Linux IO header in LibBSD. Making these conditional here means BSPs build and the imported Linux header is untouched.

Support MVME2307 DEC Tulip driver

Back port the MVME2307 patches from 6-freebsd-12 to support the MVME2307 (motorola_powerpc) NIC. The patches:

• Add nexus bus support for the ifdc NIC and UKPHY

• Update the bus space support to support BSP define split PCI address spaces

• Support PCI legacy support for the PowerPC

Count of postponed jobs is not set to zero for a newly created rate-monotonic period object (cloned)

Cloned from #4511:

rtems_rate_monotonic_get_status() returns an arbitrary number for the count of postponed jobs if it is called for a newly created period object. The count of postponed jobs should be cleared to zero at object creation.

posix_devctl() should return the errno directly not -1 and set errno

The RTEMS implementation of posix_devctl() from POSIX 1003.26 does not return error codes as required by the standard.

"Upon successful completion, posix_devctl() shall return zero; otherwise an error number shall be returned to indicate the error. The value returned via the dev_info_ptr argument is driver dependent."

It should return the errno value and not -1 and set the errno variable. The current behavior is wrong but understandable because the implementation is a wrapper for close() and ioctl() which do set errno and return -1.

bsps/i386: TSC calibration inaccurate (cloned)

Cloned from #4455:

The current implementation of the TSC calibration during startup yields inaccurate results.

There are 2 issues:

1. Rounding error:

The calibrate_tsc function calibrates for 1s by waiting for rtems_clock_get_ticks_per_second() ticks via the PIT.

The PIT has a frequency of 1193182 Hz. For a rtems tick of 1ms this yields 1193. Waiting for 1s means then waiting for 1193000 PIT ticks (instead of the full 1193182). For a 1 GHz TSC clock this would yield already ~150 kHz less then the correct TSC frequency.

2. Waiting for the first loop at begin of timer period

for (i = rtems_clock_get_ticks_per_second() * pc386_isrs_per_tick;
     i != 0; --i ) {
  /* We know we've just completed a tick when timer goes from low to high */
  then_lsb = then_msb = 0xff;
  do {                            <== First loop does not start at begin of clock period 
    READ_8254(now_lsb, now_msb); 
    if ((then_msb < now_msb) ||
        ((then_msb == now_msb) && (then_lsb < now_lsb)))
      break;
    then_lsb = now_lsb;
    then_msb = now_msb;
  } while (1);
}

This can yield deviations of several MHz for the TSC frequency. For example, for a tested CPU with a TSC frequency of 19167.1 MHz yields a calibration result of 1912.3 MHz.

Proposed solution

• Just wait 1193182 PIT ticks for 1 s instead of a number of rtems ticks
• Use the raw timer value instead of waiting for clock ticks.

libbsd/i386: Include header error through bus.h

Currently compilations of applications fails at rtemsbsd/include/bus.h:

#ifdef __i386__
  #error "your include paths are wrong"
#endif

Proposed solution: Install the header files for i386, so that the original ones from FreeBSD will be used.

clock_nanosleep() may use wrong clock for relative times (cloned)

Cloned from #4426:

For relative times, the clock identifier is not used to select the clock and instead always the CLOCK_MONOTONIC is used. A side-effect is that sleep() and nanosleep() use the wrong clock (CLOCK_MONOTONIC instead of CLOCK_REALTIME).

rtems_task_start() does not check that the entry point is not equal to NULL

Bug was introduced by commit 33829ce155069462ba410d396da431386369ed08 related to #2555.

Workspace initialization is broken for arm/imx7 and arm/raspberrypi

The ARMV7_CP15_START_WORKSPACE_ENTRY_INDEX has a wrong value. The bug was introduced by the fix for #4185.

RTEMS5: Shell: Backport commands for spi and i2c

Backport the i2c and spi shell commands of Christian Mauderer.

RTEMS5: Add spi driver for AXI SPI ip core from Xilinx (cloned)

Cloned from #4321: --> Backport to RTEMS5

Add a spidev compatible driver for the xilinx axi quad spi ip core for standard SPI mode

RTEMS5: Add driver for cadence-spi device for xilinx based BSPs (cloned)

Cloned from #4320: Backport to RTEMS5

Implement a spidev compatible driver for using the cadence-spi device of the Zynq devices in standard SPI mode.

Priority discipline is broken for semaphores and message queues in SMP configurations (cloned)

Cloned from #4358:

The priority queues in clustered scheduling configurations use a per scheduler priority queue rotation to ensure FIFO fairness across schedulers. This mechanism is implemented in the thread queue surrender operation. Unfortunately some semaphore and message queue directives use wrongly the thread queue extract operation.

rtems_semaphore_set_priority() uses an invalid SMP lock (cloned)

Cloned from #4356:

If the priority of a locked priority ceiling mutex is set, the used SMP lock sequence is invalid.

Allow RTEMS_PRIORITY for MrsP semaphores

In order to improve the compatibility of RTEMS 5.2 with future version of RTEMS which fixed #4346 allow MrsP semaphores to be created with RTEMS_PRIORITY.

grlib,grspw_pkt: SpaceWire? link-state defines corrected

Numbering of SPW_LS_CONNECTING and SPW_LS_STARTED was swapped.

grlib,l2cache: prevent unused diagnostic access

GR740 L2-cache driver improvement.

grlib,ahbstat: add new register definitions for AHBSTAT version 1

This is needed for GR740 users.

grlib,grspw_router: add function to control SpaceWire? run clock divisor

The driver needs to allow the user to set SpaceWire? run clock divisor for an individual port.

bsps,leon3: need to make cpucounter restart after underflow

Fix for the BSP timer initialization when GPTIMER is used as the cpucounter. Without restarting in HW the cpucounter would stop after some time.

bsps,leon3: _st64() did not follow the SPARC ABI

STD instruction requires even register alignment for 64-bit data.

bsps,leon3: BSP assumes that timer pre-scaler runs at 1MHz

When driver manager GPTIMER driver is used a different pre-scaler can be supported by a driver setting. This is used to increase the resolution of the timer.

bsps,leon3: avoid BSP dependency on apbuart/timer driver

The BSP amba.c references the GPTIMER and APBUART drivers causing them to be dragged in when the driver manager is used. By moving the dependencies to a separate file the user can override the default without dragging in unused drivers.

grlib,greth: added support for variable sized descriptor table sizes

grlib,grcanfd: extend the GRCAN driver with GRCANFD support

The new GRCAN_FD IP supports CAN FD standard and is mostly backwards compatible with GRCAN SW interface. It introduces support for CAN-FD frames and a new bit-rate timing registers for new baud-rate.

The GRCAN driver have been extended to support the GRCANFD IP using the same API. Device IP-core specific functions are split into different files (grcanstd.c and grcanfd.c).

grlib,can: introduce a new common CAN baud-rate timing calculating functions

There is no point in having two different implementations specific per GRLIB CAN device driver (OCCAN and GRCAN). Instead we can use the common CAN timing definitions and translate that into the device specific parameters using a new common calculation function.

grlib,ambapp: update GRLIB IP core ID list

Add the new AMBA PnP IDs of the new cores.

grlib,grspw: SET_PACKET_SIZE could use old DMA address

When the DMA table has been allocated dynamically, the IOCTL_SET_PACKETSIZE will trigger an issue where pDev->rx and pDev->tx are not updated with the new DMA tables base address. Instead the old pointers are used.

There is no point in reallocting the DMA tables because there is no configuration option to it.

grlib,gr1553bc: better check of the DMA area address alignment

GR1553 BC hardware truncates unaligned DMA memory addresses. Need to check input and the final DMA address and fail if unaligned address.

GR1553B: use new codec available in GR740 revision 1

Select the new 1553 codec available specifically in GR740.

GR1553RT driver: fixes to memory allocation and locking on descriptor updates

Issues in GRLIB GR1553B RT mode device driver (used by GR740 BSP):

gr1553rt_list_init() - fixed memory leak when dynamic allocation failed gr1553rt_bd_update() - release interrupt/spin-lock correctly

motorola_powerpc bootloader images not linking correctly

The change to support code and data sections has broken creating net boot images for the MVME 2700 (mvme2307) BSP. The linker command file ppcboot.lds does not handle the per function text sections.

Activate ehci_pci in rtems-libbsd

Import ehci_pci from freebsd-org using freebsd-to-rtems.py

fix mkimage.py script type processing

The explaination is taken from André on the user list:

Hello,

after some digging I think I found the problem and at least a workaround.

As a disclaimer, I do not know if this really counts as a general fix. I have not invested enough time to dig through the original mkimage U-Boot source to figure this out. But I have attached my workaround patch if anyone might face the same problems.

The source of the problem are eight bytes between the header and the actual input file which are missing when using mkimage.py. Also the input size and input crc are wrong which is a result of the missing bytes.

These eight bytes always reflect the actual size of the input file (first four bytes) with four bytes zeros following. Within the original mkimage tool these eight bytes are considered part of the input file section in the output file. Therefore the calculated input size is eight bytes higher and the input crc differs.

As a workaround the mkimage.py script will add these eight bytes to the output file and the input crc calculation and adjusts the input size. This only happens when the type script was selected.

Best regards André

​https://lists.rtems.org/pipermail/users/2021-January/068060.html ​https://lists.rtems.org/pipermail/users/2021-February/068140.html ​https://lists.rtems.org/pipermail/devel/2021-February/064491.html

PowerPC shared ISA IRQ support is broken (cloned)

Cloned from #4239:

The PowerPC shared ISA IRQ management in i8259s.c is broken. It fails to handle:

• Interrupt acknowledgements for the slave controller when the master has a lower priority pending interrupt acknowledge the slave and the master controllers

• The nesting support to suspend and resume an interrupt does not correctly handle the enable/disable

• The initialise of an interrupt does not handle any pending request

• Locks up when using with the IRQ server which is executing the handler in a task context

All these issues have been fixed in the i386 BSP however the code between is not shared as the i386 BSP has BSP specific IRQ management code as well.

Change motorola_powerpc bsp to support irq-generic (cloned)

Cloned from #4238:

Update the motorola_powerpc to support irq-generic moving the IRQ management to the IRQ server. This enables libbsd support for interrupts.

bsps/zynq: termios not working correctly for stdin

The termios test application does not work correctly for Zynq based BSPs. Also scanf and similar functions reading from stdin do return immediately and do not block as expected.

The console driver has been fixed in master. This should be backported to 5.x.

powerpc/motorola_shared display current RTEMS on boot

The boot messages display the RTEMS version. Display the current version and not a hard coded version:

-----------------------------------------                                                                                                                                                   
Welcome to rtems-5.0.0 (PowerPC/Generic (classic FPU)/mvme2307) on MVME 2600/2700 with MVME761
-----------------------------------------                                                                                                                    

MVME 2600/2700 has no console output (cloned)

Cloned from #4231:

A single character is output then nothing further. Tests output using the printk but spconsole01 fails.

It seems there is a problem with the UART driver interrupts.

Timeout for automatic barriers is broken (cloned)

Cloned from #4230:

A barrier wait timeout at an automatic barrier doesn't decrement the count of waiting threads.

Missing "extern" in RTEMS_LINKER_ROSET_ITEM_ORDERED_DECLARE() (cloned)

Cloned from #4223:

The fix for #3865 contains a typo. In RTEMS_LINKER_ROSET_ITEM_ORDERED_DECLARE() there is a missing "extern".

Back port rtems_interrupt_server_create() improvements

Back port 99c73303deb09ad66d4cc1c89961f62e4e618938 to RTEMS 5 to fix "fman" network interface driver issues in libbsd.

rtems_interrupt_server_delete() does not destroy the ISR lock of the server control (cloned)

Cloned from #4158:

This leads to memory corruption if RTEMS_PROFILING and RTEMS_SMP is enabled.

arm/bsps: Small MMU pages are rounded to 1 MiB (cloned)

Cloned from #4184:

The function arm_cp15_start_set_translation_table_entries rounds the end address of a section up to 1 MiB even if small 4kiB pages are used (i.e. ARM_MMU_USE_SMALL_PAGES enabled).

Will submit a patch which only rounds the end address to 4kiB boundary.

devctl.h does not compile from C++

The closing boilerplate of the C++ extern C wrapper is out of order with another #endif and the resulting code does not compile.

The attached patch addresses this but it needs to be merged to newlib and the RSB updated before this ticket closed.

Patch to newlib is on #4193.

Link problem of C++ tests when using libpci in BSP

When using libpci in a custom BSP we experienced linking problems for C++ test (e.g. spcxx01).

The reason is the missing extern C include guards for confdefs/libpci.h. Therefore, the reference to pci_config_lib_init is mangled.

A patch fixing this problem is provided here: ​https://lists.rtems.org/pipermail/devel/2020-October/062981.html

Fix NVMe disk synchronization and media block handling (cloned)

Cloned from #4164.

Improve the workaround for the LEON3FT store-store errata: TN-0009

See also:

​https://www.gaisler.com/doc/antn/GRLIB-TN-0009.pdf

Some areas were already fixed by #3057.

RFS bit map search buffer overflow

Cloned from #4148:

The bit search create writes past the end of its buffer. See:

​https://lists.rtems.org/pipermail/devel/2020-October/062701.html

rtems-bsp-tester: Off By One for Completed Jobs

Looking at ​https://lists.rtems.org/pipermail/build/2020-September/019154.html, you can see that the completed number of jobs doesn't match up with the number for the last build:

Total Time            : 2:05:13.725904 for 1696 completed job(s)
Average BSP Build Time: 0:00:04.430262

.....

 1695 profiling v850/v850sim build:
      configure: /home/joel/rtems-cron-6/rtems/configure\
      --target=v850-rtems6 --enable-rtemsbsp=v850sim --prefix=/home/joel\
      /rtems-cron-6/tools/6/bsps --enable-profiling --disable-smp

I think there are 1695 jobs completed not 1696. :)

Cortex-A9 MPCore based BSPs should include the workaround for Errata 794072 and 845369

The workaround may be already set up by the bootloader. Not having the workaround applied may lead to system crashes, so make sure the BSP does it during system startup.

RSB Missing BSP bset for xilinx_zynq_a9_qemu

This needs to be added to the RSB for 5 and master.

rtems-pairs includes invalid BSP

On both 5 and master, the list of BSP pairs includes bbxm which appears to come from the file bsps/arm/beagle/simscripts/bbxm.cfg which is obviously not a build configuration file.

arm: (families:21 bsps:57)
 altcycv_devkit               altera-cyclone-v
 atsamv                       atsam
 bbxm                         beagle
 beagleboardorig              beagle
 beagleboardxm                beagle
 beagleboneblack              beagle
 beaglebonewhite              beagle

Enhance rtems-bsps to print canonical names of bsps

For some purposes it is nice to have a list of the BSPs in the form ARCHITECTURE/BSP, which can be fed into other scripts.

shell: CRTL-U sets the cursor to the wrong position

This patch fixes a tiny bug in the command line editing of the RTEMS shell. Typing CTRL-U in the shell should remove all characters left of the cursor. After pressing CTRL-U, the current implementation does wrongly place the cursor at the end of the line instead at its beginning. To reproduce the bug, start the shell and type 'abc123' (no <RETURN>):

~/src/rtems $ qemu-system-arm -net none -nographic -M realview-pbx-a9 -m 256M -kernel build/arm/realview_pbx_a9_qemu/testsuites/libtests/dl10.exe
*** BEGIN OF TEST libdl (RTL) 10 ***                                            
*** TEST VERSION: 6.0.0.d9bdf166644f612dd628fe4951c12c6f8e94ba5f                   
*** TEST STATE: USER_INPUT                                                         
*** TEST BUILD: RTEMS_DEBUG RTEMS_NETWORKING RTEMS_POSIX_API RTEMS_SMP             
*** TEST TOOLS: 10.2.1 20200904 \                                                  
(RTEMS 6, RSB 31f936a7b74d60bda609a9960c6e1a705ba54974, Newlib a0d7982)            
RTL (libdl) commands: dl, rtl                                                      
                                                                                   
                                                                                   
RTEMS Shell on /dev/foobar. Use 'help' to list commands.                           
SHLL [/] # abc123                                                                  

Then move the cursor onto the '1' by hitting three times the <ARROW-LEFT> key. Next type <CTRL>-U:

SHLL [/] # 123                                                                     

Note that the cursor is at the end of the line (after '3') instead of correctly at the beginning (on the '1'), now. Continuing typing 'echo ' incorrectly results in the output:

SHLL [/] # 123echo 123                                                             

The patch changes this behavior so that the cursor in the second last step will be on the '1' and typing 'echo ' will then correctly reflected as:

SHLL [/] # echo 123                                                                

RSB: pkgconfig.py uses python2 specific "unicode" keyword

​https://git.rtems.org/rtems-source-builder/tree/source-builder/sb/pkgconfig.py#n218 uses the unicode keyword which is python2 specific. This problem exists on both the 5 branch and the master.

bsps/pc386: Add missing license header

bspsmp.c is missing a licence header.

License files missing on 5-freebsd-12 branch

The commits

​https://git.rtems.org/rtems-libbsd/commit/libbsd.py?id=c64a1ebaf9c8871f3f72b8754097e272043012d6 ​https://git.rtems.org/rtems-libbsd/commit/libbsd.py?id=e2fc9082816452844a53dc97f862375b74d9523b ​https://git.rtems.org/rtems-libbsd/commit/libbsd.py?id=589f43d6af2302623229760fb0f37378241b8bfe

haven't been added to the 5-freebsd-12 branch. Therefore the License files are missing.

POSIX Compliance Guide Missing Some math.h Methods

The following methods are provided as macros on top of compiler built-ins. They have been missed by previous reviews but have long been supported. The POSIX Compliance Guide for the 5 branch and master needs to be updated.

• fpclassify()
• isfinite()
• isgreater()
• isgreaterequal()
• isless()
• islessequal()
• islessgreater()
• isnormal()
• isunordered()
• nexttowardf()
• signbit()

libbsd: Back port ifmcstat command

The ifmcstat command is a useful diagnostics command to track down issues with multicasts.

RSB does not pass BSP options correctly.

Using 5.1 RTEMS release and RSB to build the beast I've found out it's not able to pass BSP options correctly using --with-rtems-bspopts option:

../source-builder/sb-set-builder
--prefix=$HOME/sfw/rtems/5.1-smp2-vga80x50-key-to-reset 5/bsps/pc
--with-rtems-tests --with-rtems-smp
--with-rtems-bspopts="BSP_VIDEO_80x50=1 BSP_PRESS_KEY_FOR_RESET=1"

command fails with:

Build Set: Time 0:00:28.975685
Build Set: Time 0:03:36.556012
installing: 5/bsps/pc ->
/export/home/karel/sfw/rtems/5.1-smp2-vga80x50-key-to-reset
clean staging: 5/bsps/pc
Staging Size: 4.271GB
Build Set: Time 0:24:53.756984
Build Set: BSP_PRESS_KEY_FOR_RESET=1
error: no build set file found: BSP_PRESS_KEY_FOR_RESET=1.bset
Build Set: Time 0:00:00.000420
Build FAILED

It looks like even if I use just one BSP options, it's still not passed to the BSP, but build is built fine at least.

arm: arm_cp15_set_translation_table_entries() should affect Inner Shareable domain

The arm_cp15_set_translation_table_entries() uses the wrong TLB maintenance operation in multi-processor systems.

RSB needs to support waf with no python command

Recent Linux distros may not provide a python command by default. This breaks building any waf based tools such as the rtems-tools and rtems-libbsd.

Provide support for waf builds by check for a suitable Python executable and using it to directly invoke waf. For example:

/usr/local/bin/python3.7 ./waf configure --prefix=/blah/blah

The reworked <rtems/confdefs.h> has a cyclic dependency with RTEMS_MULTIPROCESSING enabled

The changes for #3875 introduced a cyclic dependency between <rtems/confdefs/mpci.h> and <rtems/confdefs/threads.h>.

RSB: UnicodeDecodeError?

The RSB some times cannot create a report due to an Unicode decode error:

error: building powerpc-rtems5-gdb-9.1-x86_64-linux-gnu-1
Build FAILED
error: failure to create error report
Build Set: Time 0:01:29.712543
Traceback (most recent call last):
  File "../source-builder/sb/cmd-set-builder.py", line 26, in <module>
    setbuilder.run()
  File "/scratch/git-rtems-release/5.1-rc2/rtems-source-builder-5.1-rc2/source-builder/sb/setbuilder.py", line 736, in run
    b.build(deps, mail = mail)
  File "/scratch/git-rtems-release/5.1-rc2/rtems-source-builder-5.1-rc2/source-builder/sb/setbuilder.py", line 473, in build
    self.build_package(configs[s], b)
  File "/scratch/git-rtems-release/5.1-rc2/rtems-source-builder-5.1-rc2/source-builder/sb/setbuilder.py", line 258, in build_package
    _build.make()
  File "/scratch/git-rtems-release/5.1-rc2/rtems-source-builder-5.1-rc2/source-builder/sb/build.py", line 603, in make
    self._generate_report_('Build: %s' % (gerr))
  File "/scratch/git-rtems-release/5.1-rc2/rtems-source-builder-5.1-rc2/source-builder/sb/build.py", line 126, in _generate_report_
    self.opts, header, footer)
  File "/scratch/git-rtems-release/5.1-rc2/rtems-source-builder-5.1-rc2/source-builder/sb/ereport.py", line 56, in generate
    l.write(os.linesep.join(r))
UnicodeDecodeError: 'ascii' codec can't decode byte 0xe2 in position 50: ordinal not in range(128)
RSB Error Report: 

RSB 5/rtems-riscv fails to build on Windows

To repeat run rtems-release-test 5 1-rc2 on Windows.

C:/msys64/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: ada-tasks.o: in function `memcpy':
C:/msys64/mingw64/x86_64-w64-mingw32/include/string.h:202: undefined reference to `__memcpy_chk'
C:/msys64/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: breakpoint.o: in function `strcpy':
C:/msys64/mingw64/x86_64-w64-mingw32/include/string.h:228: undefined reference to `__strcpy_chk'
C:/msys64/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: C:/msys64/mingw64/x86_64-w64-mingw32/include/string.h:228: undefined reference to `__s
trcpy_chk'
C:/msys64/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: breakpoint.o: in function `memset':
C:/msys64/mingw64/x86_64-w64-mingw32/include/string.h:208: undefined reference to `__memset_chk'
C:/msys64/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: breakpoint.o: in function `sprintf(char*, char const*, ...)':
C:/msys64/mingw64/x86_64-w64-mingw32/include/stdio.h:366: undefined reference to `__chk_fail'
C:/msys64/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: cli/cli-decode.o: in function `strcat':
C:/msys64/mingw64/x86_64-w64-mingw32/include/string.h:234: undefined reference to `__strcat_chk'
C:/msys64/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: C:/msys64/mingw64/x86_64-w64-mingw32/include/string.h:234: undefined reference to `__m
emcpy_chk'
C:/msys64/mingw64/bin/../lib/gcc/x86_64-w64-mingw32/9.2.0/../../../../x86_64-w64-mingw32/bin/ld.exe: C:/msys64/mingw64/x86_64-w64-mingw32/include/string.h:234: undefined reference to `__m
emcpy_chk'

I think this has been resolved in a later GDB version.

NFS Client is broken on 64-bit targets

Packed and unpacked structures are mixed leading to invalid XDR encode/decode operations if addresses > 4GiB are used.

Test compilation error in riscv bsps

Configuring the riscv bsps with --enable-tests fails to build at least the dl06 test.

../rtems-5.0.0-m2006-2/configure --target=riscv-rtems5 --prefix../install/bsps --disable-networking --enable-tests --enable-rtemsbsp=griscv

Error during the compilation step:

rtems-ld -r /localdata2/somm_ja/rtems/5-0-rc/build-riscv/riscv-rtems5/c/griscv \

-C riscv-rtems5-gcc -c "-march=rv32imafd -mabi=ilp32d" \ -O rap -b dl06.pre -e rtems_main -s \ -o dl06.rap dl06-o1.o dl06-o2.o -lm

error: rap::object: Section index '0' not found: dl06-o1.o Makefile:8479: recipe for target 'dl06.rap' failed make[5]: * [dl06.rap] Error 10 make[5]: Leaving directory '/localdata2/somm_ja/rtems/5-0-rc/build-riscv/riscv-rtems5/c/griscv/testsuites/libtests' Makefile:663: recipe for target 'libtests' failed make[4]: * [libtests] Error 2 make[4]: Leaving directory '/localdata2/somm_ja/rtems/5-0-rc/build-riscv/riscv-rtems5/c/griscv/testsuites' Makefile:1222: recipe for target 'testsuites' failed make[3]: * [testsuites] Error 2 make[3]: Leaving directory '/localdata2/somm_ja/rtems/5-0-rc/build-riscv/riscv-rtems5/c/griscv' Makefile:716: recipe for target 'all-recursive' failed make[2]: * [all-recursive] Error 1 make[2]: Leaving directory '/localdata2/somm_ja/rtems/5-0-rc/build-riscv/riscv-rtems5/c/griscv' Makefile:289: recipe for target 'all-recursive' failed make[1]: * [all-recursive] Error 1 make[1]: Leaving directory '/localdata2/somm_ja/rtems/5-0-rc/build-riscv/riscv-rtems5/c' Makefile:410: recipe for target 'all-recursive' failed make: * [all-recursive] Error 1 make all 112,71s user 27,01s system 102% cpu 2:16,13 total

Gitlab mirror

I just registered the 'rtems' group over at gitlab.com.

It doesn't cost us anything to run mirrors at both Github and Gitlab when I have a chance I will setup a Gitlab mirror though it's not a priority.

I'll work out getting access to everyone through their Gitlab accounts at a later date.

Docs location for docs.rtems.org

Where do the built docs from Buildbot need to go for docs.rtems.org?

Right now I see the top is 'master' so just replace those?

Block device disk management implementation can underflow rtems_disk_device.uses

The source file discdevs.c contains logic error with regard to a rtems_disk_device.uses member, in that the .uses value can underflow and wrap-around because this member is unsigned and the implementation does not check for a current value of 0.

Case in point, see diskdevs.c line 373:

physical_disk->uses -= deleted_count;

If deleted_count is greater than physical_disk->uses, e.g. physical_disk->uses is 0 and deleted_count is 1, physical_disk->uses will underflow. There are other references to the uses member that can result in errors, e.g., line 448:

uses = --dd->uses;

We discovered this bug via a unit-test that creates multiple RAM disks and then attempts to delete them, without any uses, resulting in RTEMS errors because the uses member had wrapped around. To circumvent this bug, we had to force the uses member to a value of 1 prior to calling rtems_disk_release(), which then successfully deleted the RAM disks via rtems_disk_delete(), otherwise the results cascaded into error situations that were undesirable.

Stream of Unicode Warnings from git hook on Commit

When I pushed a commit this afternoon, I got so many of these that I can't scroll back to the git commit command:

remote: Code point 0x0000 is not Unicode, may not be portable at /data/support/git-support/hooks/hook_email.pl line 183. remote: Malformed UTF-8 character (unexpected non-continuation byte 0x03, immediately after start byte 0xfd) in print at /data/support/git-support/hooks/hook_email.pl line 183. remote: Code point 0x0000 is not Unicode, may not be portable at /data/support/git-support/hooks/hook_email.pl line 183. remote: Malformed UTF-8 character (unexpected non-continuation byte 0xfe, immediately after start byte 0xf6) in print at /data/support/git-support/hooks/hook_email.pl line 183. remote: Code point 0x0000 is not Unicode, may not be portable at /data/support/git-support/hooks/hook_email.pl line 183. remote: Malformed UTF-8 character (unexpected non-continuation byte 0x04, immediately after start byte 0xfd) in print at /data/support/git-support/hooks/hook_email.pl line 183. remote: Code point 0x0000 is not Unicode, may not be portable at /data/support/git-support/hooks/hook_email.pl lin^{CKilled by signal 2.}

How To Use SPARC memscrub

The SPARC memory scrubber code on the RTEMS master is undocumented from a using standpoint. Google shows the RCC has documentation (chapter 49) using it. Can this information be made available in the RTEMS Ecosystem somehow?

tms570: Register Macro Redefinition

This looks like the same name was accidentally used. Without knowing the architecture, you can't know what was intended:

/home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:164:0: warning: "TMS570_SYS1_CDDIS_VCLKAOFF_SET" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:163:0: warning: "TMS570_SYS1_CDDIS_VCLKAOFF_GET" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:162:0: warning: "TMS570_SYS1_CDDIS_VCLKAOFF" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:146:0: warning: "TMS570_SYS1_CSDISCLR_CLRCLKSR_OFF_SET" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:145:0: warning: "TMS570_SYS1_CSDISCLR_CLRCLKSR_OFF_GET" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:144:0: warning: "TMS570_SYS1_CSDISCLR_CLRCLKSR_OFF" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:134:0: warning: "TMS570_SYS1_CSDISSET_SETCLKSR_OFF_SET" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:133:0: warning: "TMS570_SYS1_CSDISSET_SETCLKSR_OFF_GET" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:132:0: warning: "TMS570_SYS1_CSDISSET_SETCLKSR_OFF" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:122:0: warning: "TMS570_SYS1_CSDIS_CLKSROFF_SET" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:121:0: warning: "TMS570_SYS1_CSDIS_CLKSROFF_GET" redefined /home/joel/rtems-work/rtems-testing/rtems/rtems/bsps/arm/tms570/include/bsp/ti_herc/reg_sys.h:120:0: warning: "TMS570_SYS1_CSDIS_CLKSROFF" redefined

throwing std::runtime on PC BSP fails.

Throwing a std::runtime() exception locks up.

The lock up is in the exception clean up handler where the exception object is destructed. The destructor loops distructing the std::string object. The path ends up in libstdc++-v3/include/ext/atomicity.h line 48 or exchange_and_add.

At a guess it would seem like the C++ atomics on i386 is broken or fragile.

UPDATE: This was broken when gcc i386 eliminated -mcpu in favor of -march/-mtune. The multilibs were built with -mtune and not -march.

RTEMS gdb python script do not load.

Loading the scripts does not work ...

(gdb) python import rtems
Traceback (most recent call last):
  File "<string>", line 1, in <module>
  File "/opt/work/rtems/4.12/share/gdb/python/rtems/__init__.py", line 31, in <module>
    import rtems
  File "/opt/work/rtems/4.12/share/gdb/python/rtems/rtems.py", line 38, in <module>
    import threads
  File "/opt/work/rtems/4.12/share/gdb/python/rtems/threads.py", line 39, in <module>
    import rbtrees
ImportError: No module named rbtrees
Error while executing Python code.
(gdb)