<div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div dir="ltr"><div>Hi!</div><div><br></div><div>I started to collect some more evidence on why type category-based local warnings are a worthy addition. I worked used Clang as a target of this experiment because we are all familiar with this project. The method was to grep for some keywords like temporary, dangling, StringRef, and ArrayRef. I filtered the results since some of the commits were not related to lifetime issues at all (of course, there is no guarantee that I did not miss a lot of interesting commits). After that, I compiled each commit with the modified clang and categorized the results:</div><div><br></div><div>Errors that would have been caught by the warning (so less noise on the build bots):<br>* <a href="https://github.com/llvm-mirror/clang/commit/be0345e47273a01065874e568ad673f89b75317a">https://github.com/llvm-mirror/clang/commit/be0345e47273a01065874e568ad673f89b75317a</a> (<span class="gmail-blob-code-inner gmail-blob-code-marker">StringRef RetTyName = RetTy.<span class="gmail-pl-c1">getAsString</span>();</span>)<br>* <a href="https://github.com/llvm-mirror/clang/commit/7140ad92debc63af78a9c50dcad2a0b6d4ebc680">https://github.com/llvm-mirror/clang/commit/7140ad92debc63af78a9c50dcad2a0b6d4ebc680</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/51ff8d7aa82cee89c4770832084ef5052692090a">https://github.com/llvm-mirror/clang/commit/51ff8d7aa82cee89c4770832084ef5052692090a</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/4fe9644dc3533e4517f980645957c6fc9408c6f6">https://github.com/llvm-mirror/clang/commit/4fe9644dc3533e4517f980645957c6fc9408c6f6</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/20f89390c7f9eaf7c62bcc9145fbc5ba7725557b">https://github.com/llvm-mirror/clang/commit/20f89390c7f9eaf7c62bcc9145fbc5ba7725557b</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/a638ce2afc68a2f67142ca8c7842172d49be61c0">https://github.com/llvm-mirror/clang/commit/a638ce2afc68a2f67142ca8c7842172d49be61c0</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/77cd074b2f44d342085066e3a2046e76a8d0a655">https://github.com/llvm-mirror/clang/commit/77cd074b2f44d342085066e3a2046e76a8d0a655</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/dfa25cb2d16f812bb28bdf60b6b891f4e9140072">https://github.com/llvm-mirror/clang/commit/dfa25cb2d16f812bb28bdf60b6b891f4e9140072</a><br></div><div><br></div><div>Old revision did not compile with modern Clang but the problem is very likely to be caught:</div><div>* <a href="https://github.com/llvm-mirror/clang/commit/3846ca29a8cc1d376a4b695194c29952dbbfb544">https://github.com/llvm-mirror/clang/commit/3846ca29a8cc1d376a4b695194c29952dbbfb544</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/3791814b2c5efb3986ec4dd40c75c9d28f325222">https://github.com/llvm-mirror/clang/commit/3791814b2c5efb3986ec4dd40c75c9d28f325222</a></div><div><br></div><div>Errors that are not caught because assignments are not checked, only initializations (could be added as a separate flag?):</div><div>* <a href="https://github.com/llvm-mirror/clang/commit/77c4e2380a572bfab344ca1bfc7b59beb9cf8fff">https://github.com/llvm-mirror/clang/commit/77c4e2380a572bfab344ca1bfc7b59beb9cf8fff</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/9670762c0d7fb6f3c13c61f044e335cf2f9523f9">https://github.com/llvm-mirror/clang/commit/9670762c0d7fb6f3c13c61f044e335cf2f9523f9</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/a6fb4e07ed2e9145596490240faa1eebff5a53c0">https://github.com/llvm-mirror/clang/commit/a6fb4e07ed2e9145596490240faa1eebff5a53c0</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/61ed9caca5dd6a9b9ccee7fb51296234e6cc68ad">https://github.com/llvm-mirror/clang/commit/61ed9caca5dd6a9b9ccee7fb51296234e6cc68ad</a><br></div><div><br></div><div>Errors that are not caught because they would need some semantic information about APIs that are not covered by type categories or CFG based analysis:</div><div>* <a href="https://github.com/llvm-mirror/clang/commit/3c57879f0dd98cf30cdc0672ad727f150d304c1f">https://github.com/llvm-mirror/clang/commit/3c57879f0dd98cf30cdc0672ad727f150d304c1f</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/431fb79151297df73530ca45b57f7481fc5a9bd3">https://github.com/llvm-mirror/clang/commit/431fb79151297df73530ca45b57f7481fc5a9bd3</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/9fec733e8e62b1e2ccf0d39916899f2c2d92c764">https://github.com/llvm-mirror/clang/commit/9fec733e8e62b1e2ccf0d39916899f2c2d92c764</a> (The problem is related to <span class="gmail-blob-code-inner gmail-blob-code-marker"><span class="gmail-pl-k">const</span> PrintingPolicy &Policy; member of <span class="gmail-blob-code-inner gmail-blob-code-marker"><span class="gmail-pl-en">FailedBooleanConditionPrinterHelper</span></span></span>)</div><div>* <a href="https://github.com/llvm-mirror/clang/commit/1b48633d308460049e1132960f3c33c850de5bd2">https://github.com/llvm-mirror/clang/commit/1b48633d308460049e1132960f3c33c850de5bd2</a><br></div><div><br></div><div>Errors not caught, probably bug in the prototype (and should be feasible to catch), need to investigate:</div><div>* <a href="https://github.com/llvm-mirror/clang/commit/97973546416c9d04b518c1b9572815b42872400f">https://github.com/llvm-mirror/clang/commit/97973546416c9d04b518c1b9572815b42872400f</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/1d22fa209d194c8ab676498b13ff5e25cb07d17d">https://github.com/llvm-mirror/clang/commit/1d22fa209d194c8ab676498b13ff5e25cb07d17d</a></div><div>* <a href="https://github.com/llvm-mirror/clang/commit/563fb903fab15972e2d9c66b0ae046a94be86a71">https://github.com/llvm-mirror/clang/commit/563fb903fab15972e2d9c66b0ae046a94be86a71</a><br></div><div><br></div><div>Skipped, did not see any code that could trigger the behavior while examining the diff and did not bother looking deeper:</div><div>* <a href="https://github.com/llvm-mirror/clang/commit/35276390608adec9840c7ed64f2ee211162a97cb">https://github.com/llvm-mirror/clang/commit/35276390608adec9840c7ed64f2ee211162a97cb</a><br></div><div><br></div><div>All of these commits slipped through the review process. I feel like this data could justify these checks (even more so since we can restrict it to annotated and STL types by default reducing the number of false positives to virtually 0).</div><div>What do you think?<br><br></div><div>Regards,<br></div><div>Gábor<br></div><div><br></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, 1 Mar 2019 at 14:03, Manuel Klimek <<a href="mailto:klimek@google.com">klimek@google.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Mar 1, 2019 at 1:49 PM Gábor Horváth <<a href="mailto:xazax.hun@gmail.com" target="_blank">xazax.hun@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, 1 Mar 2019 at 12:54, Manuel Klimek <<a href="mailto:klimek@google.com" target="_blank">klimek@google.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr">(correcting dmitri's email address)<br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Mar 1, 2019 at 12:24 PM Gábor Horváth <<a href="mailto:xazax.hun@gmail.com" target="_blank">xazax.hun@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, 1 Mar 2019 at 12:13, Manuel Klimek <<a href="mailto:klimek@google.com" target="_blank">klimek@google.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Mar 1, 2019 at 12:10 PM Manuel Klimek <<a href="mailto:klimek@google.com" target="_blank">klimek@google.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr">-apple George + google George to prevent bounces :)<br><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Mar 1, 2019 at 12:06 PM Manuel Klimek <<a href="mailto:klimek@google.com" target="_blank">klimek@google.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, Mar 1, 2019 at 11:51 AM Gábor Horváth <<a href="mailto:xazax.hun@gmail.com" target="_blank">xazax.hun@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><div dir="ltr">Hi Manuel,<br></div><br><div class="gmail_quote"><div dir="ltr" class="gmail_attr">On Fri, 30 Nov 2018 at 12:04, Manuel Klimek <<a href="mailto:klimek@google.com" target="_blank">klimek@google.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><br><br><div class="gmail_quote"><div dir="ltr">On Fri, Nov 30, 2018 at 12:00 PM Gábor Horváth <<a href="mailto:xazax.hun@gmail.com" target="_blank">xazax.hun@gmail.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><div><div><div><div>Unfortunately, we do not have real world experience on a large codebase for the following reasons:<br></div>* The implementation we have does not support annotations yet, and large projects are likely to contain exceptions to the default rules (even if only a few)<br></div>* The analysis has an expected coding style, e.g. it does not reason about arithmetic on raw pointers<br><br></div>We did run it on some projects like LLVM and there were a lot of noise mainly due to the unsupported language features like pointer arithmetic.<br></div><div>The rest of the false positives were mostly due to the basic assumption of the analysis that every path is feasible.<br></div><div>We did find some true positives with a similar analysis looking for use after moves in LLVM. Those true positives were redundant std::forward calls that are not going to cause any runtime error but still removing them would make the code cleaner and less likely to misbehave in the future. See: <a href="https://github.com/llvm-mirror/llvm/blob/master/include/llvm/Support/Error.h#L894" target="_blank">https://github.com/llvm-mirror/llvm/blob/master/include/llvm/Support/Error.h#L894</a> (The same arguments forwarded multiple times in a loop. Since the callee never actually moves the arguments we will not end up using a moved from object. But if we never move the argument, why would we use std::forward in the first place?)<br><br></div><div>But even if the cfg-based lifetime analysis end up being a coding style specific check, the rest of the (cfg-less) clang warnings would become much more powerful after type categories are upstreamed. They could catch a lot of errors without having any false positives by default. Thus, I do see the value going forward, even if we do not have much real world experience yet. <br></div></div></div></blockquote><div><br></div><div>Could you run the restricted checks on llvm?</div></div></div></blockquote><div><br></div><div>Sorry for the delay, but finally I found some time to actually implement the CFG-less checks and run them on LLVM. The quick prototype is available here: <a href="https://github.com/mgehre/clang/commits/lifetime-warning-typecategories" target="_blank">https://github.com/mgehre/clang/commits/lifetime-warning-typecategories</a></div><div>Literally, all of the false positives I encountered after running on LLVM and Clang were due to llvm::ValueHandleBase being miscategorized as an owner (because of its user-defined destructor).</div><div>So annotating this class would render the LLVM codebase clean of these new lifetime errors. Do you think this is a strong enough evidence to start upstreaming the type categories or do you think we should do additional measurements?</div></div></div></div></blockquote><div><br></div><div>I'd like to get Richard's opinion on this.</div></div></div></blockquote></div></div></blockquote><div><br></div><div>So just to make sure I understand correctly:</div><div>The result is that in LLVM we have:</div><div>- 1 false positive root cause (causing X false positives)</div><div>- 0 true positives</div><div>?</div><div> </div></div></div></blockquote><div><br></div><div>Correct.<br></div></div></div></blockquote><div><br></div><div>Unfortunately that doesn't sound like a really strong argument in favor - if you have something that works cleanly against clang trunk I can try to find somebody willing to try it on a bit of our internal codebase, but I can't promise anything.</div><div><br></div><div>Basically what I'd be looking for is some evidence that the false positive ratio is not too high, and for that we'd need at least a codebase containing true positives.</div></div></div></blockquote><div><br></div><div>I do not agree.</div></div></div></blockquote><div><br></div><div>Sorry if what I said came over stronger than I intended it to - with "not a strong argument in favor" I didn't mean that it's an argument against, just that it doesn't seem like overwhelming data for.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div> LLVM has good coverage and tests are regularly run with sanitizers on. Thus, I would except most of the lifetime related issues being caught by dynamic analysis in our case. I do remember, however, such problems to slip through the reviews and break the build bots occasionally.</div></div></div></blockquote><div><br></div><div>Would be interesting to have some evidence / examples for this that this check would have caught - I definitely agree that lifetime issues fall through the cracks, the question is whether those actually lead to true positives, or slip through the simple check.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div> I think, the value of those warnings are to catch those errors early, saving time for the developers. So even though if it is unlikely to find true positives in sanitizer clean builds for certain projects I still see the value. Note that, this is true for most of the similar warnings in clang. LLVM (and probably most other projects) is clean of -Wreturn-stack-address, still we do not question whether that warning is useful. </div></div></div></blockquote><div><br></div><div>Agreed. I'm really mainly looking for evidence for the decision. I'm not saying that evidence doesn't exist, but that we have to make sure to find it.</div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div><code></code><br></div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div></div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><div class="gmail_quote"><div>I also plan to look up some lifetime fixes/reverts in the commit history to check if some of those problems would have been caught by these warnings. I will share the results in a EuroLLVM technical talk.</div><div><br></div><div>P.S.:<br></div><div>Does anybody have a good idea how to get a list of buildbot breaking commits where the reason was lifetime related (other than grepping git history)?</div><div><br></div><div>Regards,</div><div>Gábor<br></div><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div class="gmail_quote"><div> </div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div dir="ltr"><div dir="ltr"><div></div><div><br></div><div>Regards,<br></div><div>Gabor<br></div><div><div><div><div><div><div><br><div class="gmail_quote"><div dir="ltr">On Thu, 29 Nov 2018 at 20:58, George Karpenkov <<a href="mailto:ekarpenkov@apple.com" target="_blank">ekarpenkov@apple.com</a>> wrote:<br></div><blockquote class="gmail_quote" style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex"><div>Thanks, the idea looks great!<div><br></div><div>The heuristic described in a talk on matching returned and accepted references looked fragile to me initially,</div><div>but after a bit of time I could not find obvious and common counterexamples which would confuse it.</div><div><br></div><div>Has the evaluation been done on any large project? E.g. LLVM itself, Chrome, Webkit, or any other large C++ codebase?</div><div>What is the false positive rate? What is the annotation burden to suppress those?</div><div>How many useful bugs were found?</div><div><br></div><div>I’m also very happy that more dataflow analyses are being implemented in Clang,</div><div>and I could help with reviews (though obviously I’m not the code owner for most of the components).</div><div><div><br><blockquote type="cite"><div>On Nov 29, 2018, at 8:02 AM, Gábor Horváth via cfe-dev <<a href="mailto:cfe-dev@lists.llvm.org" target="_blank">cfe-dev@lists.llvm.org</a>> wrote:</div><br class="gmail-m_-4912392917254527719gmail-m_-8827658777743510026gmail-m_1511282151283709290gmail-m_-3190871874678165098gmail-m_-5430317967382359601gmail-m_-51577839237342361gmail-m_-8364689291001979690gmail-m_-7504273241285550363gmail-m_7150140396152000941m_-1186342185363738695gmail-m_-7100975426903406487Apple-interchange-newline"><div><div dir="ltr"><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Hi!</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">This is a proposal to implement Lifetime Analysis [1] defined by Herb Sutter in Clang.</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Summary from the paper:</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">“This analysis shows how to efficiently diagnose many common cases of dangling (use-after-free) in C++ code, using only local analysis to report them as deterministic readable errors at compile time. The approach is to identify variables that are of generalized “Owner” types (e.g., smart pointers, containers, string) and “Pointer” types (e.g., int*, string_view, span, iterators, and ranges), and then use a local simple acyclic control flow graph (ACFG) analysis to track what each Pointer points to and identify when modifying an Owner invalidates a Pointer. The analysis leverages C++’s existing strong notions of scopes, object lifetimes, and const that carry rich information already available in reasonably modern C++ source code. Interestingly, it appears that with minor extension this analysis can also detect uses of local moved-from variables (use-after-move), which are a form of dangling.”</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">More details can be found in the paper [1] or in the CppCon keynote [3].</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Matthias Gehre and myself had been working on a prototype in Clang [2]. The changes are rather large, so we are planning to take an incremental approach to upstreaming the features should the community want to see this upstream.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><b><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Plans for upstreaming</span></b></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">1. Upstream Type Categorization</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Clang already performs statement-local lifetime analyses that would benefit from type categorization even before adding any other analysis.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">This includes annotating types as Owners and Pointers, and automatically inferring Owner or Point without annotation to minimize annotation burden.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Consider the following code example:</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">std::reference_wrapper<</span><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">const</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap"> </span><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">int</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">> get_data() {</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">    const</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap"> </span><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">int</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap"> i = </span><span style="font-size:11pt;color:rgb(9,136,90);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">3</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">;</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">    return</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap"> {i};</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">}</span></span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Unfortunately, today compilers do not warn on this case of returning a dangling reference. They do warn if we return a raw pointer or reference, but the compiler does not know that std::reference_wrapper also is a non-owning indirection. In the Lifetime analysis, this is diagnosed because std::reference_wrapper is recognized as a Pointer type.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">As a first step we would upstream the type categorization part of the analysis and make some clang warnings optionally use it. We would also upstream a set of annotations to give the users a way to fix potential false positives due to miscategorization. (This should be very rare according to our experience so far). By default, we could constrain the categorization for std types, whose semantics are known.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">2. Extensions of existing CFG-less analyses</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">2a. Initialization from temporaries</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">The goal is to detect Pointers that dangle on initialization, such as</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">std::string_view sv = “test”s;</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">By restricting the analysis to single statements, it has a low false-positive rate and can be done without building a CFG (i.e. faster).</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">2b. Return of locals</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">The goal is to detect returning Pointers to local variables, e.g.</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">std::reference_wrapper<</span><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">const</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap"> </span><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">int</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">> get_data() {</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">    const</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap"> </span><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">int</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap"> i = </span><span style="font-size:11pt;color:rgb(9,136,90);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">3</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">;</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;color:rgb(0,0,255);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">    return</span><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap"> {i};</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">}</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Similar to 2a also restricted to single statement.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">2c. Member pointer that dangles once construction is complete</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">struct X {</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">    std::string_view sv;</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">    X() : sv("test"s) {} // warning: string_view member bound to string temporary whose lifetime ends within the constructor</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">};</span></span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">2d. New of a Pointer that dangles after the end of the full-expression</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">new string_view("test"s) // warning: dynamically-allocated string_view refers to string whose lifetime ends at the end of the full-expression</span></span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">3. Intra-function analysis across basic blocks, excluding function call expressions</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Propagate point-to sets of Pointers across branches/loops intra-function, e.g. analysing</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">int* p = &i;</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">if(condition)</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">  p = nullptr;</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">*p; // ERROR: p is possibly null</span></span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">We have some CFG patches and some code traversing the CFG and propagating the analysis state. With the type categories already in place, this patch should be smaller. We could split these patches further by implementing null tracking in a separate patch.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">4. Function calls</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">auto find(const string& needle, const string& haystack) -> string_view [[gsl::lifetime(haystack)]];</span></span></div><span style="font-family:monospace,monospace"><br></span><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">string_view sv = find(“needle”, haystack);   </span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">sv[0]; // OK</span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">string_view sv = find(needle, “temporaryhaystack”);   </span></span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-family:monospace,monospace"><span style="font-size:11pt;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">sv[0]; // ERROR: sv is dangling</span></span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">This includes the following subparts.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">4a. Precondition checks</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Check that the psets of the arguments are valid at call site according to the lifetime annotations of the callee.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">4b. Postcondition checks</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Check that the psets returned from a function adhere to its advertised return/output psets.</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Rigorous checking of not just the function arguments but also the returned values is crucial part of the analysis.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">4c. Lifetimes annotations</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">The analysis gets pretty usable at this point. Most of the time the user does not need any annotations, but it is crucial to have them before a project can adapt it. For example, the user will occasionally want to explicitly state that a member function is “const as far as Lifetime is concerned” even though the function itself is not actually declared const (e.g., vector::operator[] does not invalidate any Pointers, such as iterators or raw pointers).</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">5. Implementing use after move analysis and exception support</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">These parts are not implemented yet in our prototype, but they will be useful additions for the analysis.</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><b><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Questions</span></b></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Does that make sense? What is the criteria for this work to be upstreamed? Who is willing to participate in reviewing the patches?</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Thanks in advance,</span></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">Gabor, Matthias, and Herb</span></div><br><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">[1] </span><a href="https://github.com/isocpp/CppCoreGuidelines/blob/master/docs/Lifetime.pdf" style="text-decoration:none" target="_blank"><span style="font-size:11pt;font-family:Arial;color:rgb(17,85,204);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:underline;vertical-align:baseline;white-space:pre-wrap">https://github.com/isocpp/CppCoreGuidelines/blob/master/docs/Lifetime.pdf</span></a></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">[2] </span><a href="https://github.com/mgehre/clang" style="text-decoration:none" target="_blank"><span style="font-size:11pt;font-family:Arial;color:rgb(17,85,204);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:underline;vertical-align:baseline;white-space:pre-wrap">https://github.com/mgehre/clang</span></a></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">[3] </span><a href="https://www.youtube.com/watch?v=80BZxujhY38" style="text-decoration:none" target="_blank"><span style="font-size:11pt;font-family:Arial;color:rgb(17,85,204);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:underline;vertical-align:baseline;white-space:pre-wrap">https://www.youtube.com/watch?v=80BZxujhY38</span></a></div><div style="line-height:1.38;margin-top:0pt;margin-bottom:0pt"><span style="font-size:11pt;font-family:Arial;background-color:transparent;font-weight:400;font-style:normal;font-variant-ligatures:normal;font-variant-caps:normal;font-variant-east-asian:normal;text-decoration:none;vertical-align:baseline;white-space:pre-wrap">[4] </span><a href="https://godbolt.org/z/90puuu" style="text-decoration:none" target="_blank"><span style="font-size:11pt;font-family:Arial;color:rgb(17,85,204);background-color:transparent;font-weight:400;font-style:normal;font-variant:normal;text-decoration:underline;vertical-align:baseline;white-space:pre-wrap">https://godbolt.org/z/90puuu</span></a></div><br>

</div>
_______________________________________________<br>cfe-dev mailing list<br><a href="mailto:cfe-dev@lists.llvm.org" target="_blank">cfe-dev@lists.llvm.org</a><br><a href="http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev" target="_blank">http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-dev</a><br></div></blockquote></div><br></div></div></blockquote></div></div></div></div></div></div></div></div></div>
</blockquote></div></div>
</blockquote></div></div></div>
</blockquote></div></div>
</blockquote></div></div>
</blockquote></div></div>
</blockquote></div></div>
</blockquote></div></div>
</blockquote></div></div>
</blockquote></div></div>
</blockquote></div>