Hello I had this bug since 3.60, that's why I stayed back to 3.50 for a while. I see 3.80 fixed some other bug I had, so I decided to send a report for this one.
https://www.dropbox.com/s/zko66w2asxbboqa/test01.zip?dl=0 There is a place in my code (version that works or doesn't provided) where a line like this: for (y = 0; y<y0; y++) { single_column[y] = 128; single_column[COL_MAX_HEIGHT - y - 1] = 64; } will result in assembly code that would do something like this for (y = 0; y<y0; y++) { single_column[y] = 128; single_column[COL_MAX_HEIGHT - 1] = 64; } When debugging it seemed that the IY register is loaded with start address of single_column then BC is added to it, having value equal to 'y' so IY = &single_column[y] Then IY is preserved for the second line, it's not reset back to &single_column[0] and then BC=COL_MAX_HEIGHT-1 is calculated and added to IY, resulting to single_column[y + COL_MAX_HEIGHT -y -1] = single_column[COL_MAX_HEIGHT-1] so I always fill only element COL_MAX_HEIGHT-1 in the array and not the previous ones (screenshots of right and wrong memory in the zip also) Now, main0.c is the original with the bug, main1.c is a simplified version that keeps this code but removes the other parts that are not necessary. But this removal will simplify the code generation and the bug will not occure. I don't see anything wrong in the main1.c code from my programmer's logic, and this would work properly from 3.50 and before. Things like initing few local variables as unsigned instead of signed in this code would also fix it for me (it just alters the code generation and happens to not encounter the problem but I don't know when it happens) which is fine with me to continue with my project, but I decided to report this because it could cause more problems in the future and I hope things like that will be fixed in later versions. Thanks Michael Kargas _______________________________________________ Sdcc-user mailing list Sdcc-user@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/sdcc-user
