Linker Problems On Pic18f4455

[wolahr 0]

Bug description:

I just compiled, linked and ran some code on a PIC18F4455, and noticed some unexpected behaviour in my program. On investigation I found that some of my data (a char[] array) had been stored in the 0x060-0x0ff range. On the PIC18F4455 this range contains SFRS (note details given below)

 

 

Steps to reproduce:

 

declare an array of over 80 bytes:

 

char *SomeArray= " Any of this text stored between 0x80 and 0xFF may be corrupted"

 

compile, link, run debugger and see where your data is.

 

Expected behaviour:

I expect data not to change unless it is modified by my code...

 

Is the problem 100% reproduceable:

Yes.

 

IDE version: MPLab 7.34

Compiler: BoostC

Compiler version: 6.30

Target device: PIC18F4455

OS:

 

Comments:

1 - At a guess your linker is only checking the start address for blocks of ram is valid, or only the first n bytes. Ive only seen this problem using long (200 byte) blocks.

 

2 - In my program both high and low priority interrupts are enabled

 

3 - At first glance, the data sheet suggests 0x00-0xff are all General Purpose Registers, this is not true 0x80 - 0xff are SFRs in the access window, i.e. these addresses are mapped to SFRs. See device datasheet fig 5-5 on page 66 for more details.

 

4 - Short term fix for users with similar problems:

char *DontUseThisArea[128] @ 0x80;

 

I think this is a fairly serious issue as it could cause some occasional unexplained and almost entriely undebuggable problems in peoples projects.

[Dave 0]

wolahr,

 

Steps to reproduce:

 

declare an array of over 80 bytes:

 

char *SomeArray= " Any of this text stored between 0x80 and 0xFF may be corrupted"

 

compile, link, run debugger and see where your data is.

 

Expected behaviour:

I expect data not to change unless it is modified by my code...

 

I don't see this issue.

Please provide a complete small sample program.

 

Regards

Dave

[Dave 0]

wolahr,

 

I've managed to track down the problem

It occurs when a variable straddles banked and unbanked memory.

 

This fix will be available in the next release - which should be in the next few days.

 

Thanks for reporting this issue.

 

Regards

Dave

[wolahr 0]

wolahr,

 

I've managed to track down the problem

It occurs when a variable straddles banked and unbanked memory.

 

This fix will be available in the next release - which should be in the next few days.

 

Thanks for reporting this issue.

 

Regards

Dave

<{POST_SNAPBACK}>

 

Hi dave,

 

Thanks for fixing this so quickly

 

Regards

 

Will