Hi.
I'v spent the last couple of days with a bug (caused
by myself !) that I thought was enough "silly" to be told
about in public :-)

I have an application with a 12F629, that has been working
(still under development) quite OK for some time. But then,
a few days ago, hell broke lose. All kind of problems.
Adding a few NOP's in an unrelated part of the code, could
get other parts either begin to run as it should, or just
quit working at all. Occasional resets and so on.

I finaly found it, a kind of "optimization" I shouldn't
have done. In this piece of code :

;******************************************
         ...
         ...
         movwf   ir_status
         goto    end_of_ir_bit_handle

.irbh_end     code

end_of_ir_bit_handle

         dbankif gbankadr
         btfss   flag_one_bit_received
         ...
         ...
;******************************************

I decided that the GOTO wasn't needed since it just
jumped to the next instruction anyway. So I commented it out.

Now, this is relocatable code, and since the GOTO and the
label are in two different code modules, they can end up
in totaly different places by the linker. The code just run
right into whatever was placed efter the commented GOTO.
And different code from build to build, so the whole application
was realy unpredictable ! I'v been staring into the oscope
like mad, trying to find out what the h--l is going on.

Now, *sometimes*, depending on the internal magic of the linker,
these two modules *could* end up right after each other, and
suddenly everything worked ! Well, until I removed the last piece
of debugging code :-)

Jan-Erik.

--
http://www.piclist.com hint: The PICList is archived three different
ways.  See http://www.piclist.com/#archives for details.