[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: ERROR assign: messages
From: |
W. J. van der Laan |
Subject: |
Re: ERROR assign: messages |
Date: |
Sun, 18 Apr 2021 19:38:42 +0000 |
Hello Jan,
> Sadly, this is to be expected at the moment. What happens here is that
> the size as calculated for the target of the assignment is smaller than
> the size of the rvalue.
Right. So it's an unhandled case at the moment. I'll look into it.
> > The reason I am asking is because compilation does continue, it does
> > not appear to be fatal. Still, I am worried about miscompilation.
>
> Yes. This is in the area where I tried to warn about: 64bit is not yet
> supported: these are problems that we should isolate a failing test case
> from and fix.
>
> Either the warning is off, or the size calculation is wrong; mescc's
> integer promotion rules are incomplete, I fear.
Thank you! Besides these warnings, "make" finishes successfully on the platform
now.
So my next step will be to try to get the test cases in lib/tests to pass with
mescc-generated RISC-V code. It is very likely I'm going to run into this issue
soon enough, if it is a serious problem :-)
-W