3 ms·
I love strongly worded technical arguments. I am not deeply familiar with the Modules TS, but couldn't each environment solve the module name <-> translation u
by twoodfin 9y ago
I love strongly worded technical arguments.
I am not deeply familiar with the Modules TS, but couldn't each environment solve the module name <-> translation unit mapping problem in its own way? For systems that can't keep reliable track of changes to the underlying representation (i.e., files) that would mean a fairly strict naming scheme.
The mapping doesn't have to be maintained by the compiler. C/C++ already depend on a suite of more-or-less independent tools to produce runnable software. I don't see how this problem conceptually is much worse than what a linker has to do.
- humanrebar 9y ago> I don't see how this problem conceptually is much worse... I'm asking the same question to myself. There's a bit in the article about it being slower than the status quo, but I'd like to see actual benchmarks before making pronouncements. If that hypothesis is true, then surely the TS will fail before it makes it into C++20 proper.