Fix for lldb unit tests: TestAnonymous.py, et al., when using ccache

[vharron]

Summary

This patch fixes the following unit tests when the lldb unit tests are run on ubuntu

TestAnonymous.py
TestSharedLib.py
TestSharedLibStrippedSymbols.py
TestUniqueTypes.py

The unit tests fail when lldb is configured/built with ccache or when the individual unit tests are built with ccache. Under these conditions, the "compiler" variable in lldbtest.py's getCompilerVersion() contains the ccache prefix or compiler options in addition to the compiler name. The which() function in lldbutil does not handle multiple words well and returns "None" - even though clang is embedded in the compiler name.

This patch strips off any prefixes and/or options by looking for the last word in the compiler variable that appears to be an executable and then discarding everything else.

Doug

[vharron]

Siva, can you review & test? If it's still valuable, please get reviewed by someone at Apple and submit.

"Submitted D6007 to reviews.llvm.org, a patch to fix TestAnonymous.py, et al. This patch fixes unit tests when ccache is used to build lldb, by stripping off compiler arguments from the compiler name when sending the compiler name to which in lldbtest.py." - Doug Snyder

[vharron]

from Doug Snyder

[vharron]

ccache is just not supported for now

test/lldbtest.py

         version = 'unknown'
 
         compiler = self.getCompiler()
+
+        # strip off leading spaces, prefixes (e.g. ccache, distcc, etc.) and 
+        # trailing complier options to create a single-word compiler name.
+        # otherwise, unit tests will fail if:
+        #     1) check-lldb is configured/built w/ ccache
+        #     2) they are built w/ ccache
+        compiler_split = compiler.split(" ")
+        for word in reversed(compiler_split) :
+            if which(word) != None :
+                compiler = word
+                break
+
         version_output = system([[which(compiler), "-v"]])[1]
         for line in version_output.split(os.linesep):
             m = re.search('version ([0-9\.]+)', line)