Re: Trouble building gcc-3.4.0 cygwin->linux

[Jim Wilson <wilson at specifixinc dot com>]

Dan Kegel wrote:
>[Dangit, somebody replied to this six weeks ago and told me
>what configuration variable controlled this choice, but I
>can't find the email now that I have time and need to fix this >problem!]

That would be me.  See
    http://gcc.gnu.org/ml/gcc/2004-04/msg00906.html
I just grepped for LTHUNK, found it in cp/method.c, and pointed you at 
ASM_OUTPUT_DEF which seemed to be the culprit.

Looking at it again, I see that there are some curious __CYGWIN__ ifdefs 
in there which might instead be the problem.  Your cygwin-x-linux cross 
compiler would have been compiled with a compiler that defined 
__CYGWIN__, and thus some stuff in cp/method.c would have been enabled 
and/or disabled that should not have been.  To fix this, put a
#define THIS_IS_CYWGIN_REALLY
in config/i386/cygwin.h (or maybe cygming.h?) and modify cp/method.c to 
use that define instead of __CYGWIN__.  We need a more elegant solution 
for gcc mainline, but one has already been written.  This is the 
TARGET_USE_LOCAL_THUNK_ALIAS_P stuff I mentioned earler.

See PR 14808.  This contains the method.c.diff patch that broke gcc-3.4. 
 This was an attempt to write a safer simpler patch than the one that 
went into mainline, which can be found here
    http://gcc.gnu.org/ml/gcc-patches/2004-04/msg00432.html
Unfortunately, this patch breaks cross compilers built on cygwin.

I'm not keen on trying a cygwin build to reproduce any of this, as then 
I won't be able to read mail.
--
Jim Wilson, GNU Tools Support, http://www.SpecifixInc.com