I confirm this BUG which appends on my FC31 Rpi aarch64.
I “think” 4104 can be interpreted as VT_EXTERN | VT_FLOAT (i.e 0x00001000 | 0x8)
Quickly trying following diff looks to fix this issue (and makes the whole test
suite still happy). I let maintainers tell me if it is the right fix to apply.
diff --git a/arm64-gen.c b/arm64-gen.c
index 5606271..5cd9e6b 100644
--- a/arm64-gen.c
+++ b/arm64-gen.c
@@ -1775,7 +1775,8 @@ ST_FUNC void gen_cvt_ftoi(int t)
ST_FUNC void gen_cvt_ftof(int t)
{
- int f = vtop[0].type.t;
+ int f = vtop[0].type.t & VT_BTYPE;
+
assert(t == VT_FLOAT || t == VT_DOUBLE || t == VT_LDOUBLE);
assert(f == VT_FLOAT || f == VT_DOUBLE || f == VT_LDOUBLE);
if (t == f)
From: Tinycc-devel [mailto:[email protected]] On
Behalf Of Charles Lohr
Sent: Tuesday, November 05, 2019 07:52
To: [email protected]
Subject: [Tinycc-devel] Floating/Double issue on arm64 in mob
Using default settings, in hash 96f1fb1, current mob head, I am having an
internal compiler assert on arm64. The following program causes an assert.
There are other cases but the assert remains the same.
float fd;
int main()
{
if( fd < 5.5 ) return 1;
return 0;
}
root@omitted:~/apps/cnovr/lib/tinycc# ./tcc failtest.c -Iinclude
tcc: arm64-gen.c:1781: gen_cvt_ftof: Assertion `f == VT_FLOAT || f == VT_DOUBLE
|| f == VT_LDOUBLE' failed.
Aborted
Digging a little further, turns out f is of type 4104. I'm not sure what the
best path forward here is. Any clues?
_______________________________________________
Tinycc-devel mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/tinycc-devel