https://issues.dlang.org/show_bug.cgi?id=15190
Issue ID: 15190 Summary: scope(success/failure) are oddly implemented and mask stack traces Product: D Version: D2 Hardware: All OS: All Status: NEW Severity: major Priority: P1 Component: dmd Assignee: nob...@puremagic.com Reporter: thecybersha...@gmail.com Created attachment 1558 --> https://issues.dlang.org/attachment.cgi?id=1558&action=edit Demonstration program scope statements seem to be rethrowing exceptions, or otherwise causing them to be (re)thrown them from within the stack frame of the function containing the scope statement, and not the function where the exception was thrown. This causes loss of information on the exception's origin, which makes debugging difficult. This is especially severe when the exception does not itself contain a stack trace (e.g. it's an OutOfMemoryError or an InvalidMemoryOperationError). I'm attaching a demonstration program. To reproduce the issue, compile the program (and specify one of the defined versions), start it under gdb and run "break _d_throwc", "run" and "where". Depending on the version specified, the f!int function invocations will or will not be visible in the stack trace displayed by gdb. Here are my results (Linux, DMD64 v2.068.1, gdb 7.7.1): Version | f!int visible? | Description --------+----------------+-------------------- test0 | Yes | No scope/try blocks testSS | No | scope(success) testSF | No | scope(failure) testSX | Yes | scope(exit) testTN | Yes | try/catch testTF | Yes | try/finally >From what was tested, only scope(success) and scope(failure) rethrew the exception from main(). Things of note: 1. As far as I can see, scope(success) has no reason to interfere with exceptions at all, all it should do is virtually insert code before every return point in the function. 2. scope(exit) displays the full stack trace, despite that it is simply the union of scope(success) and scope(failure). --