---- Reported by jankit@novell.com 2005-08-23 03:06:22 MST ----

In mono_class_from_typeref don't call mono_class_init as it might've been
called recursively from mono_class_init itself.

Attached is the proposed patch and two files (rt.dll & xx.cs)
To reproduce the problem:

$ mcs /r:rt.dll xx.cs

It gives

** ERROR **: pending init java.io.ObjectInputStream

aborting...
Aborted



---- Additional Comments From jankit@novell.com 2005-08-23 03:07:41 MST ----

Created an attachment (id=168401)
rt.dll : dll being referenced




---- Additional Comments From jankit@novell.com 2005-08-23 03:08:20 MST ----

Created an attachment (id=168402)
xx.cs




---- Additional Comments From jankit@novell.com 2005-08-23 05:07:49 MST ----

Created an attachment (id=168403)
Proposed patch




---- Additional Comments From miguel@ximian.com 2005-08-23 15:47:56 MST ----

Paolo, this is necessary for our mainsoft collaboration, could you
review/approve?



---- Additional Comments From lupus@ximian.com 2005-08-24 10:08:24 MST ----

Ankit, please commit, thanks.
Could you test also if ilasm allows to encode a scope typeref
where the name is itself a nested type expression?
That is, say you have a class like:
class T {
  class A {
    class B {
    }
  }
}

And see if it's possible to output a scope name of:
 "A/B"
You may need to change ilasm or PEAPI to do this and check if the
resulting assembly ius understood by peverify and the MS runtime.
In that case we want to use something like the return_nested_in()
function (and making the runtime lazily load the nested types).



---- Additional Comments From jankit@novell.com 2005-08-24 11:31:01 MST ----

lupus, PEVerify.exe doesn't seem to like "A/B", gives "Unable to
resolve token".

Commited in svn r48780.

Imported an attachment (id=168401)
Imported an attachment (id=168402)
Imported an attachment (id=168403)

Unknown operating system unknown. Setting to default OS "Other".