<div dir="ltr">There may be more things to think. I'd like to see to land with non-single-pass approach as I suggested so that it doesn't block anyone. We can revisit this later when needed.</div><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Aug 10, 2015 at 3:14 PM, Simon Atanasyan <span dir="ltr"><<a href="mailto:simon@atanasyan.com" target="_blank">simon@atanasyan.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">atanasyan added a comment.<br>
<br>
There is one more question about single-pass relocation handling. On MIPS if you call non-PIC routine from a PIC one you should create a thunk contains three instructions. We do not known the number of the thunks until we scan all relocations. So we cannot predict the output .text section's size and offsets. IIRC ARM requires some kind of thunks too. How can the single-pass relocation calculation handle this problem?<br>
<br>
<br>
<a href="http://reviews.llvm.org/D11612" rel="noreferrer" target="_blank">http://reviews.llvm.org/D11612</a><br>
<br>
<br>
<br>
</blockquote></div><br></div>