Re: libffi patches for separate debug/opt builds

[Anthony Green <green at redhat dot com>]

On 04/28/2010 02:11 PM, Dan Witte wrote:
> Actually, I think I'm gonna revise this patch; please hold off on applying for now.
>   

NP.  I'm on the road for a couple of weeks, so probably won't apply
anything 'til I get back.

Thanks for this, and all of your other patches!

AG


> Dan
>
> ----- "Dan Witte" <dwitte@mozilla.com> wrote:
>
>   
>> Anthony,
>>
>> As the title suggests. Currently libffi always builds with '-g -O2'
>> (on gcc), regardless of --enable-debug. This means a) debugging is
>> made harder; b) MSVC builds get into trouble because we can't tell it
>> to link against the debug or opt CRT. (This matters when linking
>> libffi into other libs, since every lib must use the same CRT;
>> further, end users won't have the debug CRT installed.)
>>
>> Fix is to have --enable-debug pass -DFFI_DEBUG on the compile line,
>> which msvcc.sh can translate to link against the debug CRT. (Not
>> particularly elegant, but I don't think we want to overload -g, since
>> people might manually pass that in their CFLAGS yet still want an opt
>> build.)
>>
>> This also makes it so we don't pass -g in an opt build, since doing so
>> disables a bunch of optimizations on MSVC. People will have to use
>> --enable-debug to get symbols now, or manually set -g in their CFLAGS.
>>
>> The default build will have optimization, as before.
>>
>> The other two patches do an autoreconf, and fix a MOXIE mistake that
>> caused autoreconf to bail. It's a big patch since evidently autoreconf
>> decided to out-of-line the libtool files present under m4/, which
>> looks OK to me.
>>
>> Look OK?
>>
>> Thanks,
>> Dan.
>>