On 25.07.2012 03:12, Daniel Bienvenu wrote:
> OK, this is the result of my tests SDCC with DACMAN :
> GOOD : sdcc-snapshot-x86_64-w64-mingw32-20120718-8048.zip
> FAIL : sdcc-snapshot-x86_64-w64-mingw32-20120719-8049.zip
> FAIL : sdcc-snapshot-x86_64-w64-mingw32-20120721-8050.zip
> FAIL : sdcc-snapshot-x86_64-w64-mingw32-20120724-8054.zip
> It's apparently something that was changed between build 8048 and build
> 8049, between july 18 and 19.
> And the change log says "added wassert in z80/gen.c", could this be the
> part that makes DACMAN buggy?
No. I checked the diff between the two versions, and the change to
src/z80/gen.c is really just an additional assert, something which can
result in the compiler stopping and giving an error message, but which
cannot affect the generated code.
That leaves the change to src/SDCCicode.c, which might have introduced a
new or exposed an existing bug.
You suggested that this might be a bug in array indexing, which makes
sense since the change was to multiplication-related stuff, and array
indexing results in multiplications. Still, a short testcase would help
a lot in tracking this down.