On 09/05/2012 01:48 AM, Tommi Rantala wrote:
2012/9/4 Cody P Schafer <[email protected]>:
Disable the building of libunwind-coredump except on x86_64 and x86
(where implimentations exsist).
Technically checking just the architecture would not be enough, since
there are also some OS specific bits and pieces.
We could also provide no-op implementations for the missing
architectures, similar to (as far as I can tell) the setjmp library.
This might make distro people happier, as they could package the
coredump library for all architectures unconditionally.
The one issue with doing this is that for applications which are users
of libunwind, we exchange a build error (which is preferable) for a
runtime error.
+no_force_coredump=true
+AC_ARG_ENABLE(coredump,
+ AS_HELP_STRING([--enable-coredump],[build libunwind-coredump
library]),
+ [no_force_coredump=false; use_coredump=true],
+ [no_force_coredump=false; use_coredump=false])
+
+if $no_force_coredump; then
+ case "$host_arch" in
+ x86|x86_64) use_coredump=true ;;
+ *) use_coredump=false;;
+ esac
+fi
BTW, after applying this patch, the coredump library is no longer
built by default on my x86_64 machine. I believe that AC_ARG_ENABLE()
is not used entirely correctly here.
You're correct.
The 'signature' of AC_ARG_ENABLE is
"AC_ARG_ENABLE(option-name, help-string, action-if-present,
action-if-not-present)"
I mistook action-if-not-present for action-if-disabled.
I'll fix this. Somehow.
_______________________________________________
Libunwind-devel mailing list
[email protected]
https://lists.nongnu.org/mailman/listinfo/libunwind-devel
