 |
 |
| View previous topic :: View next topic |
| Author |
Message |
kda406
Joined: 17 Sep 2003
Posts: 97
Location: Atlanta, GA, USA
|
| Maximum number of #defines? |
 Posted: Thu Apr 17, 2008 12:00 pm |
 |
|
I'm using PCH version 4.071 with a PIC18F8722 target on a gigantic (129684 bytes ROM used) application. The program is comprised of 19 header files and 17 C files. The problem I am about to describe originally happened on 4.068, and I upgraded to 4.071 but it has the same issue.
We start with a working, compiling program, that is error free. Then I add one new definition:
| Code: | | #define LOG_CONDENSATE 0x04 // Emergency shut down due to condensate overflow |
I get an error
| Quote: | | Error 53 "C:\Devel\PIC-C\Atlanta\Atlanta-RTC.c" Line 31(1,8): Expecting function name |
The error refers to a completely different system, and does not look like an error to me (and remember, it compiled before we added the new definition). Here is Line 31 from the referenced file:
| Code: | | fprintf(TERM,"Low battery! (%lu VE2)",m); |
If I simply remark the new #define out of the code, the project once again compiles successfully.
Just to cover the bases, I have tried moving the #define to a different location of the header file, I have tried changing its name altogether, and I have changed its value. In all cases the project will not compile, reporting the same error, until I remove this define. At one point I remarked out about 5K of program just to make sure we have plenty of room, and got the same results (works with define remarked, error when defined).
Since it does not seem to have to do with the name, the value, or the location, I now have to wonder if there is a maximum number of defines in PCH?
Thanks in advance!
-Kyle |
|
 |
Ken Johnson
Joined: 23 Mar 2006
Posts: 197
Location: Lewisburg, WV
|
|
| Posted: Thu Apr 17, 2008 1:13 pm |
|
|
grasping at straws here:
Try commenting out line 31
Try commenting out line 31, and/or 1 or 2 lines before/after
Try commenting out 1 or 2 lines before/after the #define
Try this #define with some other define(s) commented out.
Debugging a compiler is such fun 
Best of luck!
Ken |
|
|
horkesley
Joined: 20 Feb 2007
Posts: 48
Location: Essex UK
|
| CCS compiler limitations |
| Posted: Thu Apr 17, 2008 2:08 pm |
|
|
Hi,
This is very interesting.
I am about to start a large project and plan to use the 18F8722.
I have used the CCS compiler/ ICD-U40 on a few projects using the 18F4520 and have not been completly satisfied with the ease of use.
I have been considering different compilers and when I read about these sorts of problems it make me think I should move away from CCS.
I am not interested in debugging compilers.
I use Microsoft Studio and write in VB.NET and C++ and I dont have these problems.
I shall follow your thread with interest.
Regards and good luck
_________________
Horkesley Electronics Limited |
|
|
Ttelmah
Guest
|
|
| Posted: Thu Apr 17, 2008 2:25 pm |
|
|
Also, look at the lines immediately in front of the 'new' definition, and search the entire project for 'LOG_CONDENSATE'.
There is a common problem, which occurs particularly in CCS (but I have seen examples, even in compilers like VC++), where one change, can 'trigger' another fault, that is normally already present, but for some reason was being masked by things that followed, and then is revealed/shows up, somewhere miles away. A 'classic' example, would be (for instance), a line that lacks it's ';' some distance in front of the new definition, but happens to get correctly terminated by a semi-colon some lines latter. Changing a line in between the actual error, and the 'lucky' termination, results in the original error changing it's effect....
Best Wishes |
|
|
kda406
Joined: 17 Sep 2003
Posts: 97
Location: Atlanta, GA, USA
|
|
| Posted: Thu Apr 17, 2008 3:09 pm |
|
|
Ttelmah and the rest,
Thanks for the input. I searched the entire project and the only reference to LOG_CONDENSATE is in this header file. The project is almost 18K lines long. So far, I have been unable to find any other problems at all. At my wits end, I will post here the entire header file which contains the new definition. I cannot remark out other defines as they are required in the program. | Code: | ///////////// Includes /////////////
#include "Dev-AT24C64.c"
#include "Lib-GCSerialLog.c"
#include "Lib-Conversions.c"
///////////// Global Definitions /////////////
#define LOGSIZE DEVAT24C64A_SIZE
#define LOGADDRESS DEVAT24C64A_ADDR0
#define LOG_MALLOCFAIL 0x00
#define LOG_FREEZE 0x01
#define LOG_LUP 0x02
#define LOG_HOP 0x03
//#define LOG_CONDENSATE 0x04
#define LOG_PORESET 0x10
#define LOG_SENSORMISSING 0x11
#define LOG_TIME_CHANGE 0x20
#define LOG_FIRMWARE_UPD 0x21
#define LOG_ET_SET 0x22
#define LOG_NC_SET 0x23
#define LOG_SETTINGS_SET 0x24
#define LOG_RESET_BUTTON 0x80
#define LOG_TESTMODE 0x81
#define LOGMOD_NULL 0x00
enum logmodStartStop8 {LOGMOD_START,LOGMOD_STOP};
enum logmodET8 {LOGMOD_ET_C1,LOGMOD_ET_C2,LOGMOD_ET_S3,LOGMOD_ET_BM,LOGMOD_ET_LP,LOGMOD_ET_STRIPS};
enum logmodCycles8 {LOGMOD_NC_C1,LOGMOD_NC_C2};
enum logmodSettings8 {LOGMOD_SETTINGS_PDA,LOGMOD_SETTINGS_FACTORY,LOGMOD_SETTINGS_BIOS};
///////////// Prototypes /////////////
void AtlLOG_Init(void);
int1 AtlLOG_EntryGet (int16 mLogAddr, int8 *iMemBlock);
void AtlLOG_EntrySet(int8 iEntryType, int8 iEntryModifier=LOGMOD_NULL, void *pEntryData=NULL); |
|
|
|
PCM programmer
Joined: 06 Sep 2003
Posts: 21708
|
|
| Posted: Thu Apr 17, 2008 3:41 pm |
|
|
| What is the total number of #define statements in your program ? |
|
|
ckielstra
Joined: 18 Mar 2004
Posts: 3680
Location: The Netherlands
|
|
| Posted: Thu Apr 17, 2008 4:50 pm |
|
|
| Quote: | | I cannot remark out other defines as they are required in the program. |
I've never seen a program using all defines present. Even if you do need all the defines in your part of the program than there are the hundreds of defines in all the standard include files.
As a simple test to see if the large number of defines has reached some kind of compiler limit you could try commenting out one of the defines in a compiler supplied header file, for example in 18F8722.h file.
| Quote: | | Error 53 "C:\Devel\PIC-C\Atlanta\Atlanta-RTC.c" Line 31(1,8): Expecting function name |
Can you post line 25 to 35 of Atlanta-RTC.c? |
|
|
Ttelmah
Guest
|
|
| Posted: Fri Apr 18, 2008 2:32 am |
|
|
One thought does 'rear it's head'.
It is possible, that CCS, just happen to use the keyword 'LOG_CONDENSATE' themselves, somewhere internally. Obviously not as a 'define' (since this would then just result in a duplicate define warning), but possibly as an internal function name, inside the maths library for example.
Try just changing the name, to something like 'CONDENSATE_LOG', and see if the problem disappears.
Best Wishes |
|
|
kamputty
Guest
|
|
| Posted: Fri Apr 18, 2008 12:10 pm |
|
|
Kyle,
Create a "Hello World" program (same mcu) with lets say 1000(?) defines and see if that works. Keep adding until you blow it up (if it does). The compiler could be getting confused with "code"+"defines"...who knows!
Okay, I just tested this...
Compiler: 3.234
Simple program, just the end result from the PICWizard for a pic18f4520, no other code other then the initial wizard stuff
I created 10,000 (ten thousand) defines
| Code: |
#define kam_0 0
#define kam_1 1
#define kam_2 2
.
.
.
#define kam_9999 9999 |
I got a segfault when we are > 4500 defines...crashed; but did not get an error message... 
FWIW...
~Kam (^8* |
|
|
PCM programmer
Joined: 06 Sep 2003
Posts: 21708
|
|
| Posted: Fri Apr 18, 2008 12:20 pm |
|
|
As I think it was said earlier in the thread, look for wacky stuff like
accidentally typing a semi-colon at the end of your #define statement.
This can cause weird compiler errors. It happened recently here to
another co-worker here at the company. I had to help him find it. |
|
|
kda406
Joined: 17 Sep 2003
Posts: 97
Location: Atlanta, GA, USA
|
| It IS the compiler! |
| Posted: Mon Apr 21, 2008 9:26 am |
|
|
First of all, THANKS to all who have helped. All the recommendations are appreciated. 
Second of all, it appears to be the compiler. Although your suggestions and questions are on topic and intelligent, since I have found a key clue that points to the compiler, I will refrain from answering all of your questions.
If I remark out all of these lines, and only allow ONE to compile at a time, then I get the following results:
| Code: | #define KYLE 0x01 // <==Compiles OK
#define LOG 0x01 // <==Compiles OK
#define LOG_CONDENSATE 0x04 // FAILS!
#define LOG_CONDENSATE 0xFF // FAILS!
#define CONDENSATE 0x04 // FAILS!
#define LOG_CCSBUG 0xFF // FAILS!
#define LOG_ASSHAT 0x04 // FAILS!
#define LOG_ 0x01 // <==Compiles OK
#define LOG_C 0x01 // FAILS!
#define KYLE_KYLE 0x01 // FAILS!
#define ASSHAT 0x01 // FAILS! |
From this, I draw the following conclusions:
The name is not the problem
The underscore is not the problem
The value is not the problem It almost appears as if the PCH compiler is running out of room for the definition names; at this point any name is okay if it is 4 characters or less.
| Quote: | | Can you post line 25 to 35 of Atlanta-RTC.c? |
Sure, I put it in the initial post: | Code: | | fprintf(TERM,"Low battery! (%lu VE2)",m); |
Thanks guys,
Kyle |
|
|
kda406
Joined: 17 Sep 2003
Posts: 97
Location: Atlanta, GA, USA
|
| Even more bizarre |
| Posted: Mon Apr 21, 2008 9:41 am |
|
|
Just found another oddity. As we know, adding this line: | Code: | | #define LOG_CONDENSATE 0x04 |
causes the program not to compile, and report the error on line 31.
However, I just found quite by accident, that if I add an extra line like this: | Code: | #define KYLE 0x01
#define LOG_CONDENSATE 0x04 |
THAT WORKS! It's like it "covers up" the compiler bug. No idea why. I guess I'll be adding useless defines in the future. 
Thanks,
Kyle |
|
|
ckielstra
Joined: 18 Mar 2004
Posts: 3680
Location: The Netherlands
|
| Re: It IS the compiler! |
| Posted: Mon Apr 21, 2008 10:40 am |
|
|
| kda406 wrote: | | Quote: | | Can you post line 25 to 35 of Atlanta-RTC.c? |
Sure, I put it in the initial post: | Code: | | fprintf(TERM,"Low battery! (%lu VE2)",m); |
|
I did not ask for a single line of code but for a range of code lines, this because the CCS compiler often is wrong in pointing out the line where the actual error resides. Your line 31 looks OK, but I'd like to see the 10 lines before and 5 lines after this particular line. |
|
|
kda406
Joined: 17 Sep 2003
Posts: 97
Location: Atlanta, GA, USA
|
|
| Posted: Mon Apr 21, 2008 11:33 am |
|
|
Sorry about that. I read your request wrong even when I quoted it. Had a rough weekend, and had to come back to work to rest. 
| Code: | //--------------------------------------------------------------------------
// Do the daily maintenance. (Expand as necessary)
//
void AtlRTC_Daily(void) {
int16 m;
gsTimer.tDaily = gtNow + 86400; // Set up the daily maintenance for tomorrow
m = AtlVolt_ReadBatt(); // Get the battery voltage
if(m < VOLT_BAT_MIN) { // If it is too low, note the problem
fprintf(TERM,"Low battery! (%lu VE2)",m);
AtlTerm_CRLF();
gsLEDStatus.service = LEDON;
gbBattLow = true;
} else {
gbBattLow = false;
}
} |
I couldn't do 10 lines before because this is the first function in a library.
Thanks,
Kyle |
|
|
kda406
Joined: 17 Sep 2003
Posts: 97
Location: Atlanta, GA, USA
|
|
| Posted: Fri Jul 18, 2008 2:52 pm |
|
|
I never saw anything about this being fixed in any of the change logs, so I never upgraded. Today I upgraded to 4.076 and now I can add new #define statements. I have been coding without them for months on this project, and it has been painful.
After upgrading to 4.076, I was immediately able to add
| Code: | | #define LOG_OVERFLOW 0x8F |
with no errors. With this release, hopefully we can now close the chapter on the maximum number of defines too.
-Kyle |
|
|
|
|
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
|
Powered by phpBB © 2001, 2005 phpBB Group
|
FAQ 
Search
Register
Profile 
Log in to check your private messages 
Log in
