未捕获的异常——调试技术(C++)

发布于 2024-11-09 19:51:40 字数 634 浏览 2 评论 0原文

我遇到了一个奇怪的场景，其中以下不太可能的代码：

try{
  throw Core::ValueError();
}
catch (Core::Error &e){
  ...
}

(ValueError 继承自 Error 继承自 std::exception)

会导致如果编译到可执行文件中则捕获异常，但如果编译到特定共享库中则不会捕获异常。

所以我的问题是：

我可以使用哪些调试工具和/或技术来查看异常处理过程的黑匣子内部？我可以用gdb单步调试它吗？
我是否可以从 Mach-o 标头中提取任何信息，这些信息可以告诉我有关某些 catch 子句对某些异常的可捕获性（如果您愿意的话）的任何信息？特别是，我可以查看“gcc_except_tab”部分及其可爱的 LSDA、符号表或其他部分，并推断出符号可见性的任何问题或与捕获异常相关的其他问题吗？

我确实找到了一个在线资源，声称使用一只鸡、七个老鼠尾巴和一个粒子加速器来解决问题，但我想我应该先尝试 StackOverflow，然后将黑魔法作为最后的手段。

（我在 OSX 10.6.7 上运行 i686-apple-darwin10-g++-4.2.1）

原文

I have encountered a curious scenario in which the following unlikely code:

try{
  throw Core::ValueError();
}
catch (Core::Error &e){
  ...
}

(ValueError inherits from Error inherits from std::exception)

results in the exception being caught if compiled into an executable, but not if compiled into a particular shared library.

And so my questions:

What debugging tools and/or techniques can I use to peek inside the black-box that is the exception handling process? Can I step through it with gdb?
Is there any information I could pull out of the Mach-o headers that would tell me anything about the catchablility (if you will) of certain exceptions by certain catch clauses? In particular, can I look say at the "gcc_except_tab" section with its lovely LSDA's, or the symbols table, or another part, and deduce any problem with symbol visibility or other issue relevant to catching exceptions?

I did find an online source that claimed a solution using a chicken, seven rat tails and a particle accelerator, but I figured I'd try StackOverflow first and leave the black magic as a last resort.

(I'm running i686-apple-darwin10-g++-4.2.1 on OSX 10.6.7)

分享到QQ

分享到微博