CCS 6.1.0.00104 with Linux J-Link support v0.1.0.4 crashes when used with a EK-TM4C1294XL

Chester Gillon said:

With CCS 6.1.0.00104 running under CentOS 6.5 64-bit, was attempting to use a Segger J-Link with J-Link support v0.1.0.4 to debug a EK-TM4C1294XL.

However, CCS is repeatably crashing when attempt to start a debug session. From the progress indications prior to the crash the target is successfully programmed.

Further investigations with the same software installation and a EKS-LM4F232 board:

a) If the device is set to LM4F232H5QD in the .ccxml file then can successfully debug with the Segger J-Link under Linux.

b) If the device is set to TM4C123GH6PGE (the Tiva replacement part for the Stellaris LM4F232H5QD) then CCS crashes when try and debug using Segger J-Link under Linux.

Therefore, the crash appears sensitive to the type of device selected.

Chester Gillon said:

a) If the device is set to LM4F232H5QD in the .ccxml file then can successfully debug with the Segger J-Link under Linux.

b) If the device is set to TM4C123GH6PGE (the Tiva replacement part for the Stellaris LM4F232H5QD) then CCS crashes when try and debug using Segger J-Link under Linux.

Comparing the CCS 6.1 ccsv6/ccs_base/common/targetdb/devices/tm4c123gh6pge.xml and ccsv6/ccs_base/common/targetdb/devices/lm4f232h5qd.xml shows a difference is that following section is only in the tm4c123gh6pge.xml:

                <!-- Additional core registers -->
                <instance href="../drivers/cortexM4_fpu.xml" xml="cortexM4_fpu.xml" id="FPU" xmlpath="../drivers/" baseaddr="0x00000000" endaddr="0x00000000" size="0x00000040" accessnumbytes="4" permissions="p" />

As an experiment, commenting out the reference to cortexM4_fpu.xml from the tm4c123gh6pge.xml didn't prevent the crash using Segger J-Link under Linux.

Chester,

I apologize for the delay; I am still trying to finish an adapter to my Launchpad and see if I can reproduce this with my TM4C123 Launchpad.

However, I was able to reproduce this issue with CCSv6.1 on both Ubuntu 12.04/64 and 14.04/64 and my EK-TM4C1294XL. I am using jlink 4.96.12.

In my case, selecting the LM4F232H5QD device did not work as well - I am very surprised it worked for you.

By the way, the instance you mentioned is not significant for this particular case; it only exposes the floating-point registers of the Cortex M4F core.

I will keep investigating this and file a bug.

I apologize for the inconvenience,
Rafael

I am running into a somewhat similar issue with CCSv6.1.0 + jlink and CC2650DK. But it happens on Windows.

There was no issue on CCSv6.0.1.. until I updated the CC2650 device support to the same version that comes with CCSv6.1.0. Then the crash happened.

I wonder if they are related....

Update: It looks like a potential root cause for my issue is the GEL file that comes with the new device package. Removing the GEL file helps to make the issue go away. I haven't narrowed down which GEL action is causing the problem yet.

Rafael - yeah, I know... I was just hoping that it work anyway... oh well...

Rafael - just curious... can you try removing the startup GEL file and see if the crashes stop?

Thanks
ki

desouza said:

In my case, selecting the LM4F232H5QD device did not work as well - I am very surprised it worked for you.

Trying again with CCS 6.1.0.00104 under Linux with J-Link support v0.1.0.4 and a LM3S8962.

If simply try and start a debug session, CCS crashes (in 5 out of 5 attempts).

The alternative sequence to start a flash download and debug session seems reliable (5 out of 5 attempts have worked so far):

a) In the "Projects" folder of the Target Configurations view locate the .ccxml file for the project.

b) In the Target Configurations view right-clock on the .ccxml file and select "Launch Selected Configuration".

c) In the Debug view right click on the "Segger J-Link Emulator_0/CORTEX_M3_0" and select "Connect Target"

d) Use the Run -> Load -> Load Program menu and select the .out file for the project as the program file. This loads the program into flash and runs to main.

e) You can now successfully debug the program.

desouza said:

The fact it happens to work in certain scenarios (it works in my Windows release for certain devices) must be considered a happy coincidence...

Out of interest, is the source code for the J-Link Emulator Support available if users want to try and help debug / improve the J-Link Emulator Support, or is the source code proprietary to TI and therefore not available?