[PATCH] [libcxx] Update libc++ docs to include instructions for LIT.

Dan Albert danalbert at google.com
Fri Aug 1 19:58:42 PDT 2014


> 3. I'm interested to hear what Mac users think about this patch. The documentation seems to reflect how I build and test libc++ on linux but that may not be the case on OS X.

AIUI this works for Mac as well. It may not be what has become the typical flow for them, but this is probably more because the docs never mentioned anything other than buildit for Mac.

================
Comment at: www/index.html:107-110
@@ +106,6 @@
+  <p>
+    libc++ is known to work on the following platforms, using g++-4.2 and
+    clang (lack of C++11 language support disables some functionality). Note
+    that functinality provided by <atomic> is only functional with
+    clang.
+  </p>
----------------
Eric Fiselier wrote:
> I don't have access to a mac with g++-4.2 but is that still supported? libc++ requires a c++11 compiler to build so I'm skeptical GCC 4.2 works.
> Along the same lines I don't think the comment about lack of C++11 is correct. I think we should update this to say that a C++11 compiler is required to build libc++.
_Building_ libc++ requires a C++11 compiler (for that matter, I think it requires clang specifically). I believe this paragraph is actually about users of libc++, which is still supported for older standard versions, and from gcc.

================
Comment at: www/index.html:114-115
@@ +113,4 @@
+  <ul>
+    <li>Mac OS X i386</li>
+    <li>Mac OS X x86_64</li>
+  </ul>
----------------
Eric Fiselier wrote:
> When do you think we can say we officially support linux? 
I considered adding them. A few of the remaining test failures seem like blockers to me (though they might be innocuous, I haven't investigated).

================
Comment at: www/index.html:155-159
@@ -154,7 +166,11 @@
-  
+
+  <p>In-tree build:</p>
   <ul>
-    <li><code>cd libcxx/lib</code></li>
-    <li><code>export TRIPLE=-apple-</code></li>
-    <li><code>./buildit</code></li>
-    <li><code>ln -sf libc++.1.dylib libc++.dylib</code></li>
   </ul>
----------------
Eric Fiselier wrote:
> While I like documenting the CMake build system as the primary one since it is uniform across all systems should we be so quick to remove the documentation for the buildit script?
I was wondering this myself. It's still present in the system, though perhaps that's not obvious enough. I expect that this probably has a wide range of users, but again I wonder how many of them use it simply because it was what the docs said first. As with testit, it's also non-parallel.

Anyone have a good reason we should keep it in the docs?

http://reviews.llvm.org/D4766






More information about the cfe-commits mailing list