[cfe-dev] [analyzer] What's up with c++-allocator-inlining?

Kristóf Umann via cfe-dev cfe-dev at lists.llvm.org
Wed Feb 26 06:02:21 PST 2020


This is short and sweet. MallocChecker uses both check::newAllocator
and check::postStmt<CXXNewExpr> to model aspects of operator new. Mind that
these two are redundant not only with each other, but with the already used
check::postCall. The reason I can see is handling all values of the
analyzer config c++-allocator-inlining.

So, this flag has been true by default for a long-long time, and I
personally never changed had the need to change it. Is there a need to keep
tiptoeing around it? Here is a patch that tackles the issue, but it quite
dated and I'm not too sure about the current state of things.

[analyzer] Add a new checker callback, check::NewAllocator.

Also, shouldn't we make check::NewAllocator provide a CXXAllocatorCall
rather then a CXXNewExpr and a related under-construction SVal?

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.llvm.org/pipermail/cfe-dev/attachments/20200226/6c890603/attachment.html>

More information about the cfe-dev mailing list