On 11/15/2010 03:17 PM, Stefan Hajnoczi wrote:
On Sun, Nov 14, 2010 at 1:52 PM, Paolo Bonzini<pbonz...@redhat.com> wrote:
On 11/14/2010 02:38 PM, Andreas Färber wrote:
- --trace-file=*) trace_file="$optarg"
+ --enable-trace-file=*) trace_file="$optarg"
;;
but this should be --with-trace-file=... please. It is not being
enabled, just set to a different value.
--with-* should be reserved for library paths, but I can change it if people
prefer it that way.
Actually I think we have something similar to overriding --prefix here
:). It's a path that you can set at ./configure time.
Yeah, that's true. However...
So is it not okay to use --trace-file=<filename>?
... Autoconf would not allow unknown options not starting with --enable-
or --with-.
The rationale to avoid incompatible options in QEMU is this: suppose you
have a project using Autoconf (e.g. GCC) and you want to drop QEMU as a
subdirectory in there, e.g. to run the GCC testsuite under QEMU usermode
emulation (GCC can already do this for other simulators). To pass
options to QEMU's configure, you can include them in GCC's commandline.
The script will simply pass the option down to QEMU and it will be
processed there. However, if you pass --trace-file to GCC's configure
script, it will complain and stop.
Probably I would use something like --enable-trace-backend=simple:trace-
if I was adding something similar to an autoconfiscated project. But
unless it provides some additional benefit (as is the case with
cross-compilation support) I want to keep the syntactic changes in my
patches to the minimum.
But I know nothing of autoconf and --enable-* or --with-* sort of
make sense too.
Whatever, I have to repost the other series anyway, so I'll change to
--with-.
Paolo