Hello,
On Tue, 10 Dec 2019, Herman ten Brugge via Tinycc-devel wrote:
> I still have one small problem with testcase 111_conversion.c
>
> The problem is that it fails on i386.c (linux) when compiling with:
> tcc 111_conversion.c -o 111_conversion -b
> ./111_conversion
>
> I fixed this (and commited it in git) with the following patch in tccgen.c:
>
> if ((p->r2 & VT_VALMASK) < VT_CONST) {
> size *= 2;
> align *= 2;
> }
I've analyzed this now and pushed something slightly different with a
comment. It boils down to the above being a work-around for an
inconsistency between types and r2 entries in the vstack. My variant is
also just a work-around. Unfortunately there are many places in TCC that
aren't careful with this, so the real fixes have to wait until someone has
more time at hand. I guess adding asserts at strategic places and then
fixing them until they don't hit anymore is the simplest, but still a bit
of work. (The asserts would test that a vstack entry has a large type
that really requires two registers or has VT_CONST in r2).
Ciao,
Michael._______________________________________________
Tinycc-devel mailing list
Tinycc-devel@nongnu.org
https://lists.nongnu.org/mailman/listinfo/tinycc-devel
