On 1/9/19 5:18 AM, Segher Boessenkool wrote:
> On Wed, Jan 09, 2019 at 11:26:40AM +0100, Klaus Doldinger wrote:
>> Trying to compile head for avr-target yields the following error:
>>
>> ../../../../libgcc/libgcc2.c: In function '__negvhi2':
>> ../../../../libgcc/libgcc2.c:187:1: error: insn does not satisfy its
>> constraints:
>> 187 | }
>> | ^
>> (jump_insn 75 32 54 5 (set (pc)
>> (if_then_else (eq (zero_extract:HI (reg:QI 25 r25 [ <retval>+1 ])
>> (const_int 1 [0x1])
>> (const_int 7 [0x7]))
>> (const_int 0 [0]))
>> (label_ref 47)
>> (pc))) "../../../../libgcc/libgcc2.c":183:6 154 {*addsi3.lt0}
>> (nil)
>> -> 47)
>> during RTL pass: cprop_hardreg
>
> [ snip ]
>
> I don't see this error; how do you configure?
>
> And, please open a PR (see https://gcc.gnu.org/bugs.html ), this mailing
> list is not for bug reports. Thanks!
Note I can see this compiling gcc-7.2.0 with the 01-06-2019 snapshot.
It does not happen compiling gcc-7.2.0 with whatever gcc-8 based bits
are in Fedora 28.
That's a reasonable indication we've got a codegen issue on the trunk.
Jeff
ps. I happened to notice this as Fedora has an avr-gcc package (gcc-7.2
based) which is failing in my tester which has the 01-06-2019 snapshot
as it's buildroot compiler.
jeff
