volumit

Bitfield not correctly assigned to pointer

I can confirm your statements. So the only way is to introduce another memory qualifier for sfr direct without a type qualifier. I will implement __sfrdata (as new memory type) to solve it. I can also confirm that for __sfrdata the optimization towards __sbit like behavior is only possible in special cases. address known, etc. __data space has address range 0x00-0x7F. Any variable in it can be accessed both directly and indirectly. <- Yes. So far I remember, I detected some misleading usage of __data...

I want to do Register accesses witch are declared as bitfields. I removed the pointer discussion, because it will add just another dimension of discussion, where I would like to later follow up! Please can you check the following test case: line 31 is it done correctly and not illegal. line 248 assignment is done correctly, incl. fail on line 249 I tried as much as I can to deliver disassembly inside the testcase and comment it. I have done this not shout you, but to give you the best preparation...

It was not my intention to shout you. I think you are right, we have now several things and mixing it up. I will later on only add the initial problem description and same background info. Thanks again for your patience and taking me series.

A qemu maintainer told me by phone that I am to friendly and have to be stronger insist on my technical justifications. So again I want to contribute and not to claim, sdcc is in a good shape!

I did not claim an performance improvement for the implementation of setjmp. Just an undefined behavior according to your feedback. So I was confused, I am doing undefined, the libraries not. So fine for me, you confirmed library is also undefined. You will find more undefined test cases e.g. bug1938300.c... In addition can you confirm that the implementation of in SDCCmem.c v~~~ oid initMem () / internal stack segment ; SFRSPACE - NO FAR-SPACE - NO PAGED - NO DIRECT-ACCESS - NO BIT-ACCESS - NO CODE-ACCESS...

I did not claim an performance improvement for the implementation of setjmp. Just an undefined behavior according to your feedback. So I was confused, I am doing undefined, the libraries not. So fine for me, you confirmed library is also undefined. You will find more undefined test cases e.g. bug1938300.c... In addition can you confirm that the implementation of in SDCCmem.c v~~~ oid initMem () / internal stack segment ; SFRSPACE - NO FAR-SPACE - NO PAGED - NO DIRECT-ACCESS - NO BIT-ACCESS - NO CODE-ACCESS...

I did not claim an performance improvement for the implementation of setjmp. Just an undefined behavior according to your feedback. So I was confused, I am doing undefined, the libraries not. So fine for me, you confirmed library is also undefined. You will find more undefined test cases e.g. bug1938300.c... In addition can you confirm that the implementation of in SDCCmem.c v~~~ oid initMem () / internal stack segment ; SFRSPACE - NO FAR-SPACE - NO PAGED - NO DIRECT-ACCESS - NO BIT-ACCESS - NO CODE-ACCESS...