<div dir="ltr">Maybe I misinterpreted your previous feedback: I also don't the "check one ahead" approach in the patch.<div><br></div><div>Here's another version that uses a recursive solution like the original patch, which I think handles this much more elegantly. It's an early recurse and return so there shouldn't be much wasted extra work. What do you think?</div>
</div><div class="gmail_extra"><br clear="all"><div><span style="color:rgb(153,153,153)">-Alex</span><br></div>
<br><br><div class="gmail_quote">On Tue, Apr 1, 2014 at 9:46 AM, Jordan Rose <span dir="ltr"><<a href="mailto:jordan_rose@apple.com" target="_blank">jordan_rose@apple.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div style="word-wrap:break-word"><div>I've committed all the cleanup patches, but I'm not sure the destructor patch is correct. What if there were two temporaries within the assignment? Wouldn't we see CFGImplicitDtors for both of them? (Maybe a loop can solve this problem, although it's kind of unfortunate.)</div>
<div><br></div><div>Jordan</div><div><br></div><br><div><div><div class="h5"><div>On Apr 1, 2014, at 6:50 , Alex McCarthy <<a href="mailto:alexmc@google.com" target="_blank">alexmc@google.com</a>> wrote:</div><br></div>
</div><blockquote type="cite"><div><div class="h5"><div dir="ltr">And two more patches: the first is the refactor to split out a getRegionForConstructedObject from ExprEngineCXX.cpp: this can be applied to head.<div><br>
</div>
<div>The second fixes spurious garbage value warnings when using temporary destructor analysis. It has to be applied after both the first refactoring attached to this mail (0001-Refactor-out-a-getRegionForConstructedObject-helper-.patch) and the previous patch to add new tests (0001-Prep-work-for-fixing-C-analysis-handling-of-temporar.patch).</div>
</div><div class="gmail_extra"><br clear="all"><div><span style="color:rgb(153,153,153)">-Alex</span><br></div>
<br><br><div class="gmail_quote">On Mon, Mar 31, 2014 at 8:29 PM, Alex McCarthy <span dir="ltr"><<a href="mailto:alexmc@google.com" target="_blank">alexmc@google.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div dir="ltr">And here's a second patch, to be applied after the first, that fixes the crash while processing c++11 initializer lists. Thanks for explaining the intent of makeZeroElementRegion: I'm not sure why the FIXME I removed was there in the first place, but this seems like a much better fix than the one I had before.</div>
<span><font color="#888888">
</font></span><div class="gmail_extra"><span><font color="#888888"><br clear="all"><div><span style="color:rgb(153,153,153)">-Alex</span><br></div></font></span><div><div>
<br><br><div class="gmail_quote">On Mon, Mar 31, 2014 at 8:03 PM, Alex McCarthy <span dir="ltr"><<a href="mailto:alexmc@google.com" target="_blank">alexmc@google.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div dir="ltr">Here's a new version of the simple patch. nullptr tests are gone, and I've uncommented tests that generate incorrect warnings. As discussed, I've left a few tests commented out because they trigger crashes or asserts: I think having the tests there but commented out serves as a cheap/poor form of documentation of the current state of these bugs in clang, and it should make upcoming patches that contain fixes smaller and easier to read. Let me know if you'd like me to remove those too.<div>
<br></div><div>Since I don't have clang/llvm commit access, are you able to commit this on my behalf once this looks good to oyu?<br><br>Thanks!</div></div><div class="gmail_extra"><span><font color="#888888"><br clear="all">
<div><span style="color:rgb(153,153,153)">-Alex</span><br>
</div></font></span><div>
<br><br><div class="gmail_quote">On Mon, Mar 31, 2014 at 3:47 PM, Jordan Rose <span dir="ltr"><<a href="mailto:jordan_rose@apple.com" target="_blank">jordan_rose@apple.com</a>></span> wrote:<br><blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
<div><br>
On Mar 31, 2014, at 10:40, Alex McCarthy <<a href="mailto:alexmc@google.com" target="_blank">alexmc@google.com</a>> wrote:<br>
<br>
> Thanks for your advice, Jordan.<br>
><br>
> I've split the NORETURN printing changes to CFG.cpp and the new test additions into a separate patch (attached). Hopefully this should be a safe no-op from a behavior perspective: debug output changes slightly, and the new test cases more accurately reflects current clang behavior. Does this look safe to commit?<br>
<br>
</div>Rather than commenting out tests, I'd prefer you put them in with the "wrong" expected-warnings, and then put a FIXME next to the warnings. Commented-out tests become dead tests all too easily.<br>
<br>
I would drop the tests from nullptr.cpp. That file's for testing nullptr and nullptr_t specifically, not null pointers in general, and the new tests don't actually test any new behavior if we already believe that nullptr properly converts to a null pointer value.<br>
<span><font color="#888888"><br>
Jordan<br>
<br>
</font></span></blockquote></div><br></div></div>
</blockquote></div><br></div></div></div>
</blockquote></div><br></div>
</div></div><span><0001-Refactor-out-a-getRegionForConstructedObject-helper-.patch></span><span><0002-Fix-spurious-Undefined-or-garbage-value-returned-to-.patch></span></blockquote></div><br></div></blockquote>
</div><br></div>