Where is the program running from?

Do you mean the "blinky" program?  Are you using the StellarisWare example, or is this a new program that you made?

What tools are you using?

Hello!

I'm using CCS5.1.

Apparently there was something wrong in my installation. Last year, I managed to make
a derivative from enet_io in order to control motors and other devices through the net.
Now nothing was working anymore, so I decided to go back to the simplest possible code.
I was reading the file "Getting Started with the Cortex-M3® and StellarisWare®
Student Guide and Lab Manual" and I noticed that "import existing project" item
was missing from the project menu. I reinstalled everything and I am back on track.

But this brings me back to my question.
- I create a new program (without loading existing blinky program, just a plain new program).
- I copy blink.c into of main.c and change the path to the only include file.
Compiling and running works fine. However, if I unplug and re-plug, it does not work
anymore. What is the difference? I noticed that there are also a few c files in the blinky
project (startup,etc), so I thought that might be the reason. But it still does not work
right after plugging USB... What's the difference between importing the existing "blinky"
project and creating it from scratch by copying the files in the newly created project?

Thanks,

Pascal

The project menu "import existing project" was from CCS 4.x.  In CCS 5.x they moved that to the file menu, where you choose File-->Import, and then from there you can select the kind of thing to import, one of which is an existing CCS project.

When you create the project from scratch, do you set the part type?  Do you have a file named lm3s6965.cmd in your newly created project?  This is the linker script and it should be properly locating your program to load into flash.

My guess would be that you have a clock(crystal) problem, can you check with an oscilloscope if the signal on the crystal is oscillating? Sometimes a slightly change in capacitance(or a broken crystal) won’t let the crystal to start oscillating, but if for some reason it starts(debugging?), then it continues working.

marc_rir

clock(crystal) problem

Agree w/Marc in general - however poster states early that his is official Eval Board - thus doubt bad value/connection/device.  We have seen (and corrected) past MCU failure due entirely to excessive "C" values at xtal pins.  Again - don't believe official Eval suffers such fate.

Instead - perhaps this eval board "relies" on MCU's internal pull-ups for JTAG connections.  While JTAG probe attached - board works (suggesting that probe properly controls JTAG lines)  When probe is removed - inadequate JTAG signal levels/oscillation place the MCU into indeterminate state.  Adding external pull-ups to JTAG lines can quickly/easily confirm...  (and fix - if this is the issue)  (poster's writing somewhat unclear as to JTAG probe's presence/attachment...)

Hello!

Thanks for your replies and sorry for this mess. I finally found the solution.
Basically what happened is that the imported project (blinky) worked and
the newly created project with the same code was working only in debug mode
(i.e. not after power cycling the board). By comparing both projects, I noticed
that the file startup_ccs.c was missing in the newly created project.
Now it works fine. Sorry for all this mess. I hope this will help somebody
else in the same case.
Is there a way to automatically include this file when starting a new project?
It looks like it's exactly the same file for all the projects, and therefore
it could be automatically added.

Pascal

Thank  you for figuring out and pointing out the problem.  I'm sorry I did not spot that from the beginning.  The CCS project wizard does create the linker script but not the startup code.  I think it must be providing some default startup code from a library but it probably does not have a proper vector table.  We will work with the CCS team to see if there is a way to solve this problem.

Pascal4275

newly created project with the same code

@Stellaris Joe:

Forgive - but this NEW/USER PROJECT failure- and near variants - occurs SO often... 

Perhaps the "urge" to "create own project" should be dampened until better safeguards and/or clear/detailed written description of the process is published...