Fix bugs with name expression map APIs and test

For the API, switch to using section names instead of flags to identify
specific sections. It appears the flags set may differ across machines.

For the test, correctly set up nameExpressions and symbolNames to have
two different name expressions, one for each example in the test source
code. This fixes an invalid memory read that causes errors on some systems.
Also fix map fetch to correctly check code object instead of bitcode in second
part of test.

Change-Id: I4217f616995635515c637d38fac40303225ee6f7

Author: lamb-j

lib/comgr/src/comgr.cpp

@@ -2005,17 +2005,31 @@ amd_comgr_populate_name_expression_map(amd_comgr_data_t Data,
 
     Elf_Shdr_Impl<ELF64LE> dynsymShdr, relaShdr, rodataShdr;
     for (auto Shdr : Sections) {
-      if (Shdr.sh_type == llvm::ELF::SHT_DYNSYM)
+
+      if (Shdr.sh_type == ELF::SHT_DYNSYM)
         dynsymShdr = Shdr;
 
-      if (Shdr.sh_type == ELF::SHT_RELA)
+      // Check sh_info to differentiate .rela.dyn and not .rela
+      if (Shdr.sh_type == ELF::SHT_RELA && Shdr.sh_info == 0)
         relaShdr = Shdr;
 
-      if ((Shdr.sh_type == ELF::SHT_PROGBITS) &&
-          !(Shdr.sh_flags & ELF::SHF_WRITE) &&
-          !(Shdr.sh_flags & ELF::SHF_EXECINSTR) &&
-          !(Shdr.sh_flags & ELF::SHF_MASKPROC) )
-        rodataShdr = Shdr;
+      // We can't uniquely identify the .rodata section using the type and flag
+      // because other sections may use the exact same flags and type (i.e.
+      // .interp).  For correctness, we can check the name instead
+      if (Shdr.sh_type == ELF::SHT_PROGBITS &&
+          (Shdr.sh_flags & ELF::SHF_ALLOC)) {
+
+        Expected<StringRef> SecNameOrError = ELFFile.getSectionName(Shdr);
+        if (!SecNameOrError) {
+          llvm::logAllUnhandledErrors(SecNameOrError.takeError(),
+                                      llvm::errs(), "ELFObj creation error: ");
+          return AMD_COMGR_STATUS_ERROR;
+        }
+        StringRef SecName = std::move(SecNameOrError.get());
+
+        if (SecName.equals(StringRef(".rodata")))
+          rodataShdr = Shdr;
+      }
     }
 
     // .dynsym - Find name expressions with amdgcn_name_expr and store their
@@ -2097,9 +2111,9 @@ amd_comgr_populate_name_expression_map(amd_comgr_data_t Data,
 
     // Collect an unmangled name for each name expression
     for (auto expData : nameExpDataVec) {
-      // TODO: If/when an accessor API becomes availble to get the starting
+      // TODO: If/when an accessor API becomes available to get the starting
       // address for the section, switch to that
-      int offset = expData->RodataOffset - rodataShdr.sh_offset;
+      size_t offset = expData->RodataOffset - rodataShdr.sh_offset;
 
       // Store from the offset up until the first '\0'
       const char *unmangled =

lib/comgr/test/name_expression_map_test.c

@@ -131,8 +131,10 @@ int main(int argc, char *argv[]) {
     exit(1);
   }
 
-  char *nameExpressions[] = {"my_kernel_BOO<static_cast<int>(2+1),float >"};
-  char *symbolNames[] = {"_Z13my_kernel_BOOILi3EfEvPT0_"};
+  char *nameExpressions[] = {"my_kernel_BOO<static_cast<int>(2+1),float >",
+                             "my_kernel_FOO<static_cast<int>(2+1),float >"};
+  char *symbolNames[] = {"_Z13my_kernel_BOOILi3EfEvPT0_",
+                         "_Z13my_kernel_FOOILi3EfEvPT0_"};
 
   for (size_t I = 0; I < numNames; ++I) {
     size_t Size;