This is the mail archive of the gdb@sources.redhat.com mailing list for the GDB 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]

Re: [PATCH/RFA] sim autoconf conversion fallout


   Date: Tue, 18 Jan 2005 15:05:17 -0500
   From: Andrew Cagney <cagney@gnu.org>

   Mark Kettenis wrote:
   > There still a lot of breakage in the sim tree:

   Looks under desk, nope, no fire.  I guess MIPS, mn10300, and erc32 
   qualify as "a lot" :-)

I just tried to build a native GDB on mips64-unknown-openbsd3.6 and
sparc64-unknown-freebsd5.3 in a row.  Gave me a slightly biased view
of reality ;-).  It seems though that d10v is also broken.

   > * Several simulators no longer build because the common directory
   >   isn't configured.  AFAICT all similators need that directory, so I
   >   propose the attached patch.

   This problem applied to just two simulators:  MIPS and mn10300.  I've 
   fixed the error and confirmed that the MIPS builds.

Thanks for fixing it.  However, what's against configuring the common
directory unconditionally?  Makes things even simpler ;-).

   > * The --enable-sim option is essentially a no-op now, since all
   >   simulators are built unconditionally if gdb is configured for a
   >   matching target.  Why has the option been retained while all the
   >   logic that used it is gone?  Why was the logic removed in the first
   >   place?

   --enable-sim always was a no-op (unlike --disable-sim).   The only 
   exception was for erc32 and since that doesn't build at all I've now 
   disabled it.

Ah, I overlooked the "if test "${enable_sim}" != no".

   > * In the past several simulators were only built when compiling with
   >   gcc.  That code was ripped out.  Why?

   Why not.

    From '98 (although I suspect '96):
      # The PowerPC simulator uses the GCC extension long long as well as
      # ANSI prototypes, so don't enable it for random host compilers
      # unless asked to.
   hence the test.

   Fast forward 7 years - even GCC now assums ISO-C and long-long is a 
   defacto "standard".

With erc32 gone this seems not unreasonable.  If "long long" poses a
problem we can always bring over some of the logic from bfd.

Mark


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