I need help with this unknown build problem

[PCM programmer]

Read this thread about trying different optimization levels (with #opt):
http://www.ccsinfo.com/forum/viewtopic.php?t=23991

This is a two-page thread. Read it and my post at the end
about "don't do anything weird":
http://www.ccsinfo.com/forum/viewtopic.php?t=24548

Read the links in this thread, which are about "random resets".
Some of the solutions may help with your problem.
http://www.ccsinfo.com/forum/viewtopic.php?t=27638&start=4

Another thread about trouble-shooting strange problems:
http://www.ccsinfo.com/forum/viewtopic.php?t=19721

[JimmyNg]

PCM programmer,
Thank you so much for the links. I read them and tried some of the sugesstions, but the problem won't go away. I have not tried the #opt level yet, I will try it soon. By the way, I use the 18F8720 PIC

The problem is so unexpected, just adding in some more dummy codes or printf statements, might make the problem disappeared.

Even with the problem, runing in debug mode seems OK. In the Non debugger mode, It seems that the build can be loaded on the target, but during the power resets, it can't start unless I hold the MCLR pin to gound for a second( I do have a 47 Kohm pullup at the MCLR pin)...

I really want to undersand the root cause of this problem, but in case if i can't find out. I wonder if I add some dummy code to make the problem goes away, then how reliable the build is and can i trust the build to use it for release????

Jim

[PCM programmer]

Are you sure that you're programming the PIC for "standalone" mode ?

Here is a thread that describes how to do that.
http://www.ccsinfo.com/forum/viewtopic.php?t=28356&highlight=standalone+icd

Also, look for anything that's strange in your hardware connections.
For example, sometimes a board will only work if the ICD-U40 is
connected to it. That's because the ICD-U40 was providing the
ground connection (it was left off the design by accident).
http://www.ccsinfo.com/forum/viewtopic.php?t=28072&highlight=icd+ground

[JimmyNg]

hi,
I tried with the #opt 7....The code size jumps up from 39% to 47%. The problem goes away. I wonder if later on, I add more code into the build, then the problem will be appreare again.

With higher level 8->11. It won't work.

I like to know more about the #opt. Choosing the higher OPT, will make the compiler procudes smaller the code side at the expense of execution time?

[PCM programmer]

Compile it with #opt 7 and #opt 8. Compare the differences between
the .LST files. Look for something that could cause the problem.