Test suite failures for clang x86-64 and Sparc

Hello everyone,

Just wondering if CodeGen/mandel.c and CodeGen/array.c are supposed to
fail from the testsuite (maybe there is a list of possible/probable fail
cases?).

I am running a Linux x86-64 with svn updates to llvm and clang today.
Clang outputs 'invalid operands to binary operator' for the mandel.c
failure and a 'segmenation fault' for array.c. I can publish full logs
if you like.

Clang is using the 64 bit libs (/lib64/libc.so.6), by the way.

I can look into this a little more if needed. I just checked bugzilla
and there doesn't seem to be bugs listed for this.

Thanks,
K.Wilson

P.S. I just checked clang on my SPARC/Linux machine and only
CodeGen/array.c fails. It is an internal 'bus error' in this case with
no stack trace. This one has no diagnostic or logging info at all :frowning:

CodeGen/mandel.c should print an error message - Unimplemented aggregate expr. However, neither test should cause 'segmentation fault' or 'bus error'. Please file bugzilla reports along with preprocessed source files.

Yep, unfortunately, since mandel.c includes system headers, it's success can vary based on what the system headers include. Please add a .i file (clang -E mandel.c > mandel.i) to bugzilla and we can investigate, thanks!

-Chris

Hello everyone,

Just wondering if CodeGen/mandel.c and CodeGen/array.c are supposed to
fail from the testsuite (maybe there is a list of possible/probable fail
cases?).

Nope, nothing should fail. I'm curious why array.c fails, it doesn't pull in any system headers.

I am running a Linux x86-64 with svn updates to llvm and clang today.
Clang outputs 'invalid operands to binary operator' for the mandel.c
failure and a 'segmenation fault' for array.c. I can publish full logs
if you like.

stack traces would be very useful.

I can look into this a little more if needed. I just checked bugzilla
and there doesn't seem to be bugs listed for this.

Ok

P.S. I just checked clang on my SPARC/Linux machine and only
CodeGen/array.c fails. It is an internal 'bus error' in this case with
no stack trace. This one has no diagnostic or logging info at all :frowning:

Note that mandel.c is mostly ifdef'd out for sparc machines. Back in the day when I first was working with it, solaris was lacking proper c99 support, so GCC wouldn't build it.

-Chris

Kelly,

Yep Devang,

That did it. Array.c now compiles properly on x86-64 and Sparc/Linux.
Nice catch.

Thanks,
K.Wilson

P.S. Should I update the bug (#1760) or does someone else do that?

Yep Devang,

That did it. Array.c now compiles properly on x86-64 and Sparc/Linux.
Nice catch.

Thanks,

no problem :slight_smile:

K.Wilson

P.S. Should I update the bug (#1760) or does someone else do that?

go ahead.