[PATCH] D48084: [FileSpec] Delegate common operations to llvm::sys::path

[Pavel Labath via Phabricator via llvm-commits]

labath accepted this revision.
labath added inline comments.


================
Comment at: source/Utility/FileSpec.cpp:244-246
+  // Only update style if explicitly requested.
+  if (style)
+    m_style = (*style == Style::native) ? GetNativeStyle() : *style;
----------------
JDevlieghere wrote:
> labath wrote:
> > I don't really have a clear opinion on whether the new behaviour here is better or not, but it any case this sounds like something worth explicitly calling out.
> As far as the FileSpec class is concerned this doesn't change anything, we were always explicitly passing the m_style argument which is now implicit. To me this seems strictly better than what is was, were you might accidentally end up with the host path. 
> 
> I'll add a comment in the header.
Yes, but the counterargument to that is: since you're replacing the old filespec with a completely unrelated path, who's to say that the old path syntax is any better fit than "host".

I think it would be best here to actually make this argument mandatory and force everyone to think about the correct syntax when constructing the object, but maybe that's a change for another day.


================
Comment at: unittests/Utility/FileSpecTest.cpp:284
     "//",
-    "//a",
     "//a/",
----------------
JDevlieghere wrote:
> labath wrote:
> > Any particular reason for removing this?
> Yup, llvm doesn't consider `C:` or `//net` to be absolute paths (as opposed to `C:\` or `//net/`. While this is debatable, it makes things consistent with drive names on Windows, which is why I added the net tests above. 
Ah, I see.. I can certainly understand why `c:` wouldn't be considered absolute, but I'm not sure the same thing applies to `//net`. However, I think I can live that interpretation right now..


Repository:
  rL LLVM

https://reviews.llvm.org/D48084