This looks like a good improvement for clarity. Should wait for Rui's approval before submitting.

Updated comment formatting.

Changed some comments.

Removed some last unnecessary changes.

Some formatting fixes.

Don't leak symbol into assembler, remove enum class, use createTempSymbol.

Put synonymous header first.

Reformat.

Reformat

Exclude checksums

This looks good to me.

Format

Remove errant errors.

Restructrue, create real symbol not temp.

Add checksum offset directive.

Why do we have to add placeholders for the checksum offsets? Aren't they just added in order, one after another?

Fortunately this seems to work whether generating assembly or object

Make this entire thing easier to follow and understand.

Everything works finally.

Imprved?

Use MCFixup

Some more testing changes.

Add line to end of file.

tests

Use stringref

return string instead of buffer

Make isAvailable() function instead of using flag.

In D37240#861561, @ruiu wrote:

Hmm, now it doesn't look like it is towards a right direction.

As I understand the API of your WindowsManifestMerger class, it provides an empty implementation if libxml2 is not available so any code that uses your WindowsManifestMerger class can compile at least on any condition (though it may raise an error/returns an empty result at runtime if it wasn't compiled with libxml2). This patch uses conditional compilation to avoid compiler errors, which seems a violation of the rule.

Have internal lld functions return result directly and handle errors themselves.

Sorry, previous patch did not compile and method of assigning config flag was incorrect.

In D37321#859384, @vitalybuka wrote:

This seems better than using XML_PARSE_NODICT, because we also address the issue of any other memory I might have failed to copy....I could do this in a subsequent patch.

So I'll submit the patch?
I'll try to enable this on our asan bot to detect problems in future.

In D37240#859371, @zturner wrote:

We have similar issues in the PDB code, where we might want to use the native pdb reader or the DIA reader. To handle this, we have an enumeration:

/// Specifies which PDB reader implementation is to be used.  Only a value
/// of PDB_ReaderType::DIA is currently supported, but Native is in the works.
enum class PDB_ReaderType {
  DIA = 0,
  Native = 1,
};

Then, when you actually call the function to load the PDB, it looks like this:

  // Create the correct concrete instance type based on the value of Type.
  if (Type == PDB_ReaderType::Native)
    return NativeSession::createFromPdb(Path, Session);

#if LLVM_ENABLE_DIA_SDK
  return DIASession::createFromPdb(Path, Session);
#else
  return make_error<GenericError>("DIA is not installed on the system");
#endif

Then the user could write something like:

if (auto EC = loadDataForPDB(PDB_ReaderType::DIA, Session)) {
  if (auto EC2 = loadDataForPDB(PDB_ReaderType::Native, Session)) {
    return make_error<Foo>("PDB Unsupported");
  }
}

// Now I have either a native session or a DIA session, but I don't know (or care) which.
return Session;

Add flag for manifest library being enabled.

In D37321#859285, @vitalybuka wrote:

In D37321#859258, @ecbeckmann wrote:

Sorry for being so late to respond to this, I've been busy with team match this week.

Thanks for catching these errors. It was my fault for just using unique_ptr instead of explicitly calling xmlFree on the xmlDoc structures. However, I'm confused about the XML_PARSE_NODICT flag, how will this help? As far as I can tell it prevents the creation of a new string dictionary for that xmlDoc that is parsed? Which could be a problem if I had href's and namespaces in one tree point to another. However I never do this and always duplicate the entire string from one tree to another.

As soon as I start to delete OutputDoc correctly xmlFreeDoc(Doc) in ~WindowsManifestMergerImpl() fails. I guess I see this both with or without asan. XML_PARSE_NODICT helps.

Also out of curiosity how did you discover the presence of memory problems? I've looked on msan and nothing seems to be there.

check-llvm under msan disables libxml (and other 3rd party deps) as msan needs them instrumented as well.
I am using check-llvm with asan

Sorry for being so late to respond to this, I've been busy with team match this week.

Remove one more consume.

Remove unnecessary consume because toString consumes error.

Throw errors that internal mt has if they are not related to lack of libxml2.

Remove unnecessary error move assignment.

Add -allow-empty

Rebase

Add comment

Looks fine to me.

Make factory function instead of having constructor fail.

Remove unnecessary braces

Looks fine to me.

This looks fine now.

This looks fine to me.