This is the mail archive of the java@gcc.gnu.org mailing list for the Java project.


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

libffi/i386 vs. -fno-pcc-struct-return


11 libffi test cases have been failing on FreeBSD/i386.  The reason
appears to be this configuration (to mirror the system compiler):

/* Don't default to pcc-struct-return, we want to retain compatibility with
   older gcc versions AND pcc-struct-return is nonreentrant.
   (even though the SVR4 ABI for the i386 says that records and unions are
   returned in memory).  */
#define DEFAULT_PCC_STRUCT_RETURN 0

I found the failure by stepping through cls_1_1byte.exe and noting the
assumptions and then verifying that running the cases with
-fpcc-struct-return appears to fix the problem.

It seems like the correct fix is to add a new variant of the SVR4 ABI
which covers this case.  Is there a libffi maintainer willing to fix
it, or would you like me to submit a patch?  BTW, it would seem that
the best fix would track the actual compiler flags in use.  Advise?

Regards,
Loren


Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]