Re: [3.0 branch] libjava testsuite, i686-pc-linux-gnu --disable-shared

[Bryce McKinlay <bryce at waitaki dot otago dot ac dot nz>]

Andrew Haley wrote:

> This problem is caused by initialization in libgcj: class
> initialization routines are being called before the C++ static
> initializers that they depend on have been run.
>
> This patch removes the dependence on static initializers.  It is
> mostly the same as one posted by Bryce a couple of months ago.
>
> In practice it is almost impossible to control C++ static
> initializers, so I propose that we ban them from now on.

Agreed.

> +   finit_leg_name = _Jv_makeUtf8Const ("$finit$", 6);

I think we can remove this?

> RCS file: /cvs/gcc/gcc/libjava/java/lang/natClassLoader.cc,v
> retrieving revision 1.32
> diff -p -2 -c -r1.32 natClassLoader.cc
> *** natClassLoader.cc   2001/04/28 01:39:15     1.32
> --- natClassLoader.cc   2001/05/03 15:23:32
> *************** void
> *** 424,438 ****
>   _Jv_RegisterClasses (jclass *classes)
>   {
> !   static bool init = false;
> !
> !   if (! init)
> !     {
> !       init = true;
> !       _Jv_InitThreads ();
> !       _Jv_InitGC ();
> !       _Jv_InitializeSyncMutex ();
> !     }
> !
> !   JvSynchronize sync (&ClassClass);
>     for (; *classes; ++classes)
>       {
> --- 424,429 ----
>   _Jv_RegisterClasses (jclass *classes)
>   {
> !   // FIXME
> !   //JvSynchronize sync (&ClassClass);
>     for (; *classes; ++classes)
>       {

_Jv_RegisterClasses needs to be made thread-safe before we can check it in (thats
the reason for the FIXME).

regards

  [ bryce ]