Sorry, missed the question.<div><br></div><div>Pretty much. Mostly that you could just hand craft the particular expressions that you're trying to assemble rather than working up an entire llc -> .o file path. For example the kind of tests done in test/MC/COFF/secrel-variant.s.<div>
<br></div><div>Make sense?</div><div><br></div><div>-eric</div></div><br><div>On Wed Dec 18 2013 at 9:07:40 PM, Timur Iskhodzhanov <<a href="mailto:timurrrr@google.com">timurrrr@google.com</a>> wrote:</div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div dir="ltr"><div><br><div>2013/12/17 Timur Iskhodzhanov <span dir="ltr"><<a href="mailto:timurrrr@google.com" target="_blank">timurrrr@google.com</a>></span><br>
<blockquote style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><div>> OK, some ideas on how to break this up to make it a bit easier:<br>
><br>
> a) assembler support - the stuff to basically assemble various different .s files that happen<br>
> to have the format. Test cases can be simple .s files and use llvm-mc and llvm-readobj to test.<br>
> At least the label diff doesn't even really need to be a line table I wouldn't think.<br>
<br>
</div>Can you please clarify what needs to be tested using .s files?<br>
<br>
I think you want me to test that<br>
llc -filetype=asm | llvm-mc -filetype=obj<br>
generates the same file as<br>
llc -filetype=obj<br>
does for the new directives emitted by EmitCOFFSectionReference and<br>
EmitCOFFStaticOffset (namely ".secrel32" and ".offset from L1 to L2").<br>
<br>
Just wanted to double-check that I understood you correctly.</blockquote></div><br></div></div><div dir="ltr"><div>ping</div></div>
</blockquote>