This is the mail archive of the mailing list for the Java project.

Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]
Other format: [Raw text]

Re: Patch: FYI: small test suite change


On Wed, 2003-07-09 at 22:57, Tom Tromey wrote:
> Anyway, this makes one of the tests a little more efficient, so we
> don't time out when running it in the interpreter.
> from  Jeff Sturm  <>
> 	* libjava.lang/ (run): Cache .class value.

This test has been very important in the past since it showed two
1. That .class constructs were not cached at all.
2. That when doing a class lookup (and stack walk) using the interpreter
was horribly slow (which seem to have been fixed in 3.4).

Combined these problems made SyncTest fail on a lot of machines (and it
still fails on my slow powerpc box).

I agree that the purpose of SyncTest shouldn't be to test whether or not
the compiler generates (efficient) code to correctly cache .class
constructs. But shouldn't there now be a test that does explicitly test
this? That test can then probably XFAIL since gcj -C does indeed not
cache .class constructs (jikes for example does this correctly).



Index Nav: [Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav: [Date Prev] [Date Next] [Thread Prev] [Thread Next]