<div dir="ltr"><div class="gmail_extra"><br><div class="gmail_quote">On Mon, Mar 23, 2015 at 9:54 PM, Ben Langmuir <span dir="ltr"><<a href="mailto:blangmuir@apple.com" target="_blank">blangmuir@apple.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class=""><blockquote type="cite"><div>On Mar 23, 2015, at 5:38 PM, Chandler Carruth <<a href="mailto:chandlerc@google.com" target="_blank">chandlerc@google.com</a>> wrote:</div><br><div><div dir="ltr">Digging this thread back up, I'm starting to work on deterministic output of modules. This is *really* important. Unfortunately, it's almost impossible to make progress as long as this code is actually firing because it kind of makes things definitively non-deterministic. ;]</div></div></blockquote><blockquote type="cite"><div><div dir="ltr"><div><br></div><div>Ben, do you have any concerns about only emitting the signature record for implicit modules? This would omit it for explicit modules (where I'm working), PCH files, etc. Based on your commit message and explanation, it doesn't really make sense outside of implicit modules AFAICT, and we're hoping it will go away completely.</div></div></div></blockquote><div><br></div></span><div>SGTM although I would prefer we continue to check the /expected/ signature when an AST file imports an implicit module (and in particular when a PCH imports an implicit module). Thanks for working on this!</div></blockquote><div> </div><div>I'm not planning to change the checking code path at all. All the tests pass if I just restrict the emission to be when generating implicit modules.</div><div><br></div><div>Perhaps the confusion is that we have two totally different 'implicit/explicit' things in Modules land. There are implicit vs. explicit submodules. That's not what I'm talking about. I'm talking about whether -fno-implicit-modules is passed. When the build system is produces modules files for us, it doesn't seem like we ever need to take responsibility for the signature and can just omit it.</div><div><br></div><div>At least, all the regression tests pass when I make the above change, so even if it isn't the perfect state, I think it is a strict improvement. It at least allows me to test the removal of some other bit of non-determinism.</div><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><span class=""><br><blockquote type="cite"><div><div dir="ltr"><div><br></div><div><br></div><div>Also, I have one question here which may just be my ignorance:</div><div><div class="gmail_extra"><br><div class="gmail_quote">On Thu, Oct 23, 2014 at 11:05 AM, Ben Langmuir <span dir="ltr"><<a href="mailto:blangmuir@apple.com" target="_blank">blangmuir@apple.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">+ ASTFileSignature ExpectedSignature,<br>
+ std::function<ASTFileSignature(llvm::BitstreamReader &)><br>
+ ReadSignature,<br></blockquote><div><br></div><div>Wow, a std::function? Is this really necessary? Seems super heavy weight.</div></div></div></div></div></div></blockquote><div><br></div></span><div>Oops, totally unnecessary. Strength-reduced to a function pointer in r233050.</div><span class=""><br><blockquote type="cite"><div><div dir="ltr"><div><div class="gmail_extra"><div class="gmail_quote"><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex"><br>
+static ASTFileSignature readASTFileSignature(llvm::BitstreamReader &StreamFile){<br>
+ BitstreamCursor Stream(StreamFile);<br>
+ if (Stream.Read(8) != 'C' ||<br>
+ Stream.Read(8) != 'P' ||<br>
+ Stream.Read(8) != 'C' ||<br>
+ Stream.Read(8) != 'H') {<br>
+ return 0;<br>
+ }<br></blockquote><div><br></div><div>This is truly magical. I assume this is checking for some magic string? Comments or something would be really nice here. Sharing the code would be even more nice.</div></div></div></div></div></div></blockquote><div><br></div></span><div>Yeah, this is the AST/PCH file magic number. Added “startsWithASTFileMagic” to factor this out of the three places we check this, also in r233050.</div><span class=""><br><blockquote type="cite"><div><div dir="ltr"><div><div class="gmail_extra"><div class="gmail_quote"><div> </div><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
+<br>
+ // Scan for the CONTROL_BLOCK_ID block.<br>
+ if (SkipCursorToBlock(Stream, CONTROL_BLOCK_ID))<br>
+ return 0;<br></blockquote><div><br></div><div>Why would it ever be reasonable to call this for a module file without such a block?</div></div></div></div></div></div></blockquote><div><br></div></span><div>It’s not, but that’s exactly what returning 0 means - I added a doc comment to that effect. Or was there something else here?</div></blockquote><div><br></div><div>If this is a programming invariant, it should be assert()-ed rather than returning something.</div></div></div></div>