简体繁体 English

使用不同优化级别编译的不同翻译单元中的模板实例化

[英]Template instantiation in different translation units compiled with different optimization levels

原文 2016-03-14 01:59:17 3 1 c++/ templates/ gcc/ linker/ template-instantiation

Suppose I have two translation units, both of which use std::string . 假设我有两个翻译单元，都使用std::string 。 I compile one of them with -O3 , and the other without optimizations, then link the result together. 我用-O3编译其中一个，而没有优化则编译另一个，然后将结果链接在一起。 Both object files would contain instantiated std::string , but I would expect one version to be compiled with optimizations, and the other one without. 两个目标文件都将包含实例化的std::string ，但是我希望一个版本可以进行优化编译，而另一个版本则不进行优化。 Would the linker pick only one of those during linking? 链接程序在链接过程中只会选择其中之一吗？ If so, which one? 如果是这样，哪一个？ Will the resulting executable be guaranteed to always run properly? 是否可以保证生成的可执行文件始终正常运行？

Update: as this looks to be implementation-specific, it would make sense to narrow this down to gcc and clang behavior, since they both implement well-defined and concrete Itanium ABI, rather than trying to approach this from the point of view of the C++ standard itself. 更新：由于这看起来是特定于实现的，因此将其缩小到gcc和clang行为是有道理的，因为它们都实现了定义明确且具体的Itanium ABI，而不是试图从实现的角度来解决这个问题。 C ++标准本身。

1 个解决方案

As far as the C++ standard goes, this is probably as way out in far left field as it possibly can. 就C ++标准而言，这可能是在最左侧的领域中可能的解决之道。 For starters, the C++ standard has nothing to say about "optimization levels", or any of that. 对于初学者来说，C ++标准没有任何关于“优化级别”的说法。

So, this falls entirely within the scope of "implementation defined". 因此，这完全属于“定义的实现”的范围。 The answer depends on the exact compiler/linker and the compiler/linker version you're using. 答案取决于确切的编译器/链接器以及您使用的编译器/链接器版本。 The correct answer for one compiler or linker will be applicable only to that compiler or linker. 一个编译器或链接器的正确答案仅适用于该编译器或链接器。 A different compiler, or even a different version of the same compiler, can produce different results. 不同的编译器，甚至同一编译器的不同版本，都可能产生不同的结果。

I will expect one of two results: 我期待以下两个结果之一：

A) The linker will complain about merging non-identical segments, or A）链接器会抱怨合并不同的段，或者

B) One or the other will be picked at random. B）一个或另一个将被随机挑选。 Probably whichever first or last translation unit is passed to the linker. 可能是第一个翻译单元或最后一个翻译单元被传递到链接器。

So, in conclusion, the only way to know the answer to this question is for you to try it with whatever compiler or linker you're using, and examine the results. 因此，总而言之，知道该问题答案的唯一方法是让您使用正在使用的任何编译器或链接器进行尝试，然后检查结果。