RE: [Mono-list] RFC: Corlib Unit Test on Linux How-To

[Cristian Diaconu]

>>>>>>>>>>>>>>>> I'm not sure why they chose to do this, but it does look intentional. My only guess so far is that they wanted to allow developers the opportunity to inspect the program's state with a debugger before any of the finally blocks ran. <<<<<<<<<<<<<<<<   That, and the fact that having a "two-pass" stack walk matches Windows' SEH better than the single pass you suggest with finally clauses invoked as you go. In fact, I'm not even sure how the single pass would work in Mono in the presence of user-filtered handlers. (The way I read the specification, user-filtered handlers must run before any finally block is being run. Moreover, in a future release they seem to intend to support EXCEPTION_CONTINUE_EXECUTION returned from filters which would *require* the stack to be intact.)   My 2c anyway   Cristian