[PATCH] D35755: [Solaris] gcc toolchain handling revamp

[Fedor Sergeev via Phabricator via cfe-commits]

fedor.sergeev added a reviewer: tstellar.
fedor.sergeev marked 11 inline comments as done.
fedor.sergeev added inline comments.


================
Comment at: lib/Driver/ToolChains/Gnu.cpp:1840
+    // Yet, still look for RHEL devtoolsets
+    // (should it be done Linux-only??)
+    Prefixes.push_back("/opt/rh/devtoolset-6/root/usr");
----------------
aaron.ballman wrote:
> This is a good question to get an answer to before committing. :-) I don't know the answer myself, unfortunately.
So we need somebody that could answer it! :)
Last person who touched these paths was Tom Stellard.
Adding as a reviewer.


================
Comment at: lib/Driver/ToolChains/Solaris.cpp:208
   if (GCCInstallation.isValid()) {
-    GCCVersion Version = GCCInstallation.getVersion();
-    addSystemInclude(DriverArgs, CC1Args,
-                     getDriver().SysRoot + "/usr/gcc/" +
-                     Version.MajorStr + "." +
-                     Version.MinorStr +
-                     "/include/c++/" + Version.Text);
-    addSystemInclude(DriverArgs, CC1Args,
-                     getDriver().SysRoot + "/usr/gcc/" + Version.MajorStr +
-                     "." + Version.MinorStr + "/include/c++/" +
-                     Version.Text + "/" +
-                     GCCInstallation.getTriple().str());
+    const auto &Callback = Multilibs.includeDirsCallback();
+    if (Callback) {
----------------
aaron.ballman wrote:
> Shouldn't use `auto` here because the type is not spelled out in the initialization.
I believe this is the case described in LLVM coding standard as "when the type would have been abstracted away anyways". It would be MultilibSet::IncludeDirsFunc, which IMO does not add anything to readability.


https://reviews.llvm.org/D35755