Great!

still lgtm

Would you mind formatting the commit message to wrap at 80 chars? I know llvm commits are bit all over the place on the style used but I find wrapped commit message easier to read myself.

Nice. I like that we were able to delete almost as much code as was added here.

rebase

This change does point to perhaps some refactoring that we should do at some point to try to remove some the boiler place that is common among symbol types, and input elements, as well as how the symbol are added to the symbol table.

I suppose the synthetic table is there so it can one day be used for the indirect function table? Should we wait until then to actually add it?

OK, -DCMAKE_BUILD_WITH_INSTALL_RPATH=ON worked for me

This change broke the emscripten SDK builders which build with LLVM_LINK_LLVM_DYLIB and expect the resulting binaries to be portable/relocatable.

rebase

lgtm .. assuming we do eventually figure out how to trigger this using a unit test. I generally agree with you that it rather strange that ::markLive is called at all on an undefined synbol.

wasm changes lgtm assuming other backends agree.

Added a test. Once again I wish we had a good disassembler to make the test more readable.

• feedback

IIUC comdat is a propoerty shared by a group of a definitions (not declarations) so I don't see who an import could be part of a comdat group.

( i didn't see any response the the two comments, one regarding the extra debugging output and the other regarding the include the .type foo,@function lines in test code).

revert

Lets just go with duplicating the function for now (along with a TODO to followup).

revert part

Great!

One side effect is that that defs in this file that use classes from the non-atomic memory file such as LoadPatImmOff and LoadPatImmOff will no longer have the HasAtomics as part of their Requires. Does this seems reasonble?

I'm not sure how to address the duplication of got and the associated FIXME. Adding some others to the review list who might know a way to deal with that without duplication.

add test

The alternative to this approach looks like implementing something
like PredicateControl in Mips.td where we can split the predicates
into groups so they don't clobber each other.

• fix start

feedback

• rebase + add test

Still need to add testing..

In D92348#2431841, @aardappel wrote:

Well, these tests are not going to tell us if the generated instructions make any sense, but ok..

• rebase

In D91418#2431400, @rksharma wrote:

In D91418#2431314, @MaskRay wrote:

I think changing from main to to avoids an issue and adds a new issue. We should fix the issues all together, and this requires consensus from others ports to proceed.

Is the new issue, the clash with lld::*::link and if so, then ideally we should replace link from all the ports to linkerMain?

• remove some debugging

Do all these change essentially match the ELF linker? A brief look makes me thing they do.

Maybe there are two logical parts we can separate here:

I guess we wait for D92359 and then land this on top of that.

Oh I see .. up until now we import the table.

I believe there is some coverage of createInitMemoryFunction in the tests.. although maybe not in a very useful/readable way.