<html><head></head><body style="word-wrap: break-word; -webkit-nbsp-mode: space; -webkit-line-break: after-white-space; "><div><div>On Oct 25, 2012, at 1:39 PM, Lang Hames wrote:</div><blockquote type="cite"><div>Tim Northover pointed out that there's a flaw in my recent FP_CONTRACT patch (r166383). That patch restricted FP_CONTRACT (or rather tok::annot_pragma_fp_contract) to appearing as the first token in a compound statement. The problem with my patch is that it prohibits things like:</div>
<div><br></div><div><font face="courier new, monospace">void foo() {</font></div><div><font face="courier new, monospace"> #pragma MS_STRUCT ON</font></div><div><font face="courier new, monospace"> #pragma STDC FP_CONTRACT ON</font></div>
<div><font face="courier new, monospace"> // ...</font></div><div><font face="courier new, monospace">}</font></div><div><br></div><div>Which isn't prohibited by the C99 spec, and I think seems reasonable.</div><div>
<br></div><div>A simple solution to the problem is to loop at the start of a compound statement parsing tokens that can reasonably appear there (FP_CONTRACT, MS_STRUCT, maybe others?), then break out of that look when an unhandled token encountered. Attached is a patch that implements this approach, processing FP_CONTRACT and MS_STRUCT pragmas at the start of compound statements before falling through.</div>
<div><br></div><div>Not being a language lawyer, is what else should be handled in this loop? </div></blockquote><div><br></div>Other pragmas, maybe. You could potentially check for all the pragma annotation tokens.</div><div><br><blockquote type="cite"><div>Does this seem like a reasonable approach?</div></blockquote><br></div><div>It seems fine.</div><div><br></div><div>John.</div><br></body></html>