Clang analyzer run #4051
Clang analyzer run #4051
Conversation
Otherwise we'll NULL-dereference in git_attr_cache__init
1c687db to
49be45a
Compare
|
Thanks, @tiennou ! |
|
As a side note, after silencing the warnings about alloc failures and some other things, I still see 47 analyzer reports. Also, I have a question : the analyzer is "confused" by the differing stance we can have w.r.t to negative/not-zero error checks. Since I already derailed this PR with a discussion about error handling, I'll complete my tour with this one : is there a current, recommended idiom for |
Yes to both of your questions. 😝 The current, recommended idiom is: Which - is pretty deficient, as you point out, and as we've had lingering at #2716. The new recommended idiom should probably be: Which is more succinct, readable and handles positive exit codes. That's my opinion, anyway, but I haven't led the charge to make a fix. |
This is a few low-hanging fruits from using Xcode "Analyze" against libgit2.
Current
masterhas 85 of those, this one is 81, and I managed to get it down to 64 by having more fine-grained control overGITERR_CHECK_ALLOChelps. If you feel that's a worthy goal, I can make another PR with those changes.There are a few harder things which I can't tell if they're false-positives or not.