Re: Test Failues with libffi

[Anthony Green <green at redhat dot com>]

"Brent L. Bates" <blbates@vigyan.com> writes:

>      I'm trying to install the lastest version of libffi (3.0.9) on a SGI
> running IRIX 6.5.22f and the MIPS Pro 7.4.4m compilers.  The test report
> says I have 15 `unexpected failures'.  5 tests are run 3 times with different
> compiler optimization options, so I really have only 5 failures.
>      One of the failures makes no sense at all.  It says the output does not
> match what is expected, but it does match.  Here is the output:
>
> Running /usr/local/src/libffi/libffi-3.0.9/testsuite/libffi.call/call.exp ...
> FAIL: libffi.call/cls_dbls_struct.c -O0 output pattern test, is 1.0 2.0
> 1.0 2.0, should match 1.0 2.0?
> 1.0 2.0?
>
>      The only thing I can think of is there is some kind of carriage
> return/line feed mismatch as the numbers match.

What about the "?"?  Where does that come from?


>
>      The next 2 failures I don't know what to make of.  Here is their output:
>
> FAIL: libffi.call/cls_double_va.c -O0 output pattern test, is 0.0
> res: 4
> 7.0
> res: 4, should match 7.0?
> res: 4?
> 7.0?
> res: 4
> FAIL: libffi.call/cls_longdouble_va.c -O0 output pattern test, is 0.0
> res: 4
> 7.0
> res: 4, should match 7.0?
> res: 4?
> 7.0?
> res: 4

If you look at those two test cases, what happens when you put the "\n"
at the end of the dg-output string comments instead of the beginning?


>
>      The final 2 failures are because the test insists on using the C
> compiler instead of the C++ comiler on C++ source files.  Here are the errors:
>
> Running /usr/local/src/libffi/libffi-3.0.9/testsuite/libffi.special/special.exp ...
> FAIL: libffi.special/unwindtest.cc -LANG:std -D__STDC_HOSTED__ -I/usr/local/include -I/usr/local/include/openssl -OPT:Olimit=0 -TARG:processor=r10000:platform=ip28 (test for excess errors)
> WARNING: libffi.special/unwindtest.cc -LANG:std -D__STDC_HOSTED__ -I/usr/local/include -I/usr/local/include/openssl -OPT:Olimit=0 -TARG:processor=r10000:platform=ip28 compilation failed to produce executable
> FAIL: libffi.special/unwindtest_ffi_call.cc -LANG:std -D__STDC_HOSTED__ -I/usr/local/include -I/usr/local/include/openssl -OPT:Olimit=0 -TARG:processor=r10000:platform=ip28 (test for excess errors)
> WARNING: libffi.special/unwindtest_ffi_call.cc -LANG:std
> -D__STDC_HOSTED__ -I/usr/local/include -I/usr/local/include/openssl
> -OPT:Olimit=0 -TARG:processor=r10000:platform=ip28 compilation failed
> to produce executable

Not sure what to do about this.  Does SGI's C compiler driver have a
command line option to tell it that we really want to compiler C++, not
C?  That would help.

>      I had to make changes to 2 files to get rid of some compiler
> errors/warnings.  These are the changes:
>
> diff -bwir libffi-3.0.9.org/testsuite/libffi.call/ffitest.h libffi-3.0.9/testsuite/libffi.call/ffitest.h
> 18c18
> < #define CHECK(x) !(x) ? abort() : 0
> ---
>> #define CHECK(x) if( !(x) ) { abort(); }
>
>
> diff -bwir libffi-3.0.9.org/testsuite/libffi.call/nested_struct9.c libffi-3.0.9/testsuite/libffi.call/nested_struct9.c
> 33c33
> <   result.y = b2.b + b3.x.b + b4.e;
> ---
>>   result.y = (unsigned char)(b2.b + b3.x.b + b4.e);
>
>       Any help in resolving these issues would be greatly appreciated.  Thanks.

Out of curiosity, what were the errors/warnings?

Thanks!

AG