C++: Fix cpp/new-free-mismatch false positives#3206
Conversation
|
The QLDoc check is failing because this is exposing the I think moving the matching into the models is a good idea, but I might have the model make the decision about specific functions or expressions rather than using strings. In any case, we should probably talk about it in an issue rather than a PR. |
|
I've added QLDoc to the top of those files. It's fairly minimal, but points users to the files in |
|
qlformat failed :-/ |
rdmarsh2
left a comment
rdmarsh2
left a comment
There was a problem hiding this comment.
LGTM once the qlformat is fixed.
|
Sorry about that. Should pass this time... |
jbj
left a comment
jbj
left a comment
There was a problem hiding this comment.
Robert has reviewed, and the tests are passing.
3 participants
There were four new false positives for
cpp/new-free-mismatching/abseil/abseil-cpp. The issue was that calls toallocator newwere being misrecognized asmalloc-style allocations because they were function calls.I've added test cases and tidied some things up as I went along, including deprecating another part of the legacy wrapper
alloc.qll(something I've been meaning to get around to).I'm thinking about the next step being wrapping the concept of an allocation
kind("malloc","new","new[]") into the models themselves so that we can simplify this and similar code. Perhaps we should also create additionalkinds for some of the more exotic allocators, so that we can detect more kinds of mismatches.