Re: Can't build the latest snapshot with gcc-2.95.2-1.

[Chris Faylor <cgf at cygnus dot com>]

On Thu, May 18, 2000 at 12:42:59PM -0500, Mumit Khan wrote:
>On 19 May 2000, Kazuhiro Fujieda wrote:
>
>> I can't build the latest snapshot properly with gcc-2.95.2-1 as
>> the following.
>> 
>> gcc -o mount.exe mount.o -lnetapi32 -ladvapi32
>> mount.o(.text+0x29):mount.cc: undefined reference to `muto::~muto(void)'
>> /Home/fujieda/cygwin/snap/OBJ/i686-pc-cygwin/winsup/cygwin/libcygwin.a(libccrt0.o)(.text+0x29):libccrt0.cc: undefined reference to `muto::~muto(void)'
>> collect2: ld returned 1 exit status
>> make[1]: *** [mount.exe] Error 1
>> make[1]: Leaving directory `/Home/fujieda/cygwin/snap/OBJ/i686-pc-cygwin/winsup/utils'
>> make: *** [utils] Error 2
>
>This is being triggered by the defintion of __mbuf in the inline 
>definition of sigthread::init() (sigproc.h). At this point I don't
>know where the problem is, but it has to do with the fact that
>__mbuf is `static __attribute__((section(".data_cygwin_nocopy")))'.
>
>> As far as I look into the output of `nm libccrt0.o' (attached
>> below), the compiler seems to put the unnecessary reference to
>> the destructor into the libccrt0.o. Should I use the latest
>> snapshot of GCC?
>
>GCC snapshots produce the same behaviour. It's probably a bug in GCC
>-- once you start using section attributes and all that, things pretty
>hairy in C++.  We need to find a reasonably workaround.

It's interesting that this happened once I moved the initialization from
one class to another.

cgf