Re: [BC] test results from a darwin build.

[Andrew Haley <aph at redhat dot com>]

Bryce McKinlay writes:
 > Andrew Haley wrote:
 > 
 > >Andrew Haley writes:
 > > > Andreas Tobler writes:
 > > >  > Hi all,
 > > >  > 
 > > >  > just a fyi. I posted my first results from a bc build on darwin.
 > > >  > 
 > > >  > http://gcc.gnu.org/ml/gcc-testresults/2004-11/msg00718.html
 > > >  > 
 > > >  > A few ICE's and a few ': error: final field 'CONST' may not have been 
 > > >  > initialized'  (and similars) cross my eyes on a first view.
 > > >  > 
 > > >  > I rebuild to make the test a bit faster but I keep the log around.
 > > >  > 
 > > >  > Maybe someone is interested.
 > > > 
 > > > Very!  Darwin shouldn't really be worse than Linux with these tests.
 > >
 > >Ah, forget that.  I have been informed that static initializers in gcj
 > >DSOs don't work in Darwin.  Which probably explains it all...
 >
 > From what I understand, they dp work when the dylibs are explicitly
 > linked (-l), but not when they are dlopen'ed. (Otherwise, a shared
 > libgcj wouldn't work at all on Darwin). So, I don't think that the
 > shared-library initializers issue has much to do with this.

Well, perhaps not, but a gcj that works with the BC ABI and shared
libraries really needs this to work.  Therefore, until this
fundamental problem is fixed I can't see any purpose in pursuing any
Darwin failures.

Andrew.