Dan Hecht has posted comments on this change. Change subject: Add a build flag for the undefined behavior sanitizer, aka "ubsan". ......................................................................
Patch Set 2: > > If they're relying on twos-complement we could just cast to > > unsigned for the bit-manipulation, right? > > > > Are any of these overflows legitimate use of signed overflows? > > > Assuming not, shouldn't we just fix the code that can lead to > > > signed overflows? Otherwise, while the behavior would become > > > defined, it's still not correct. > > > > What is your definition of a legitimate use of a signed overflow? > > > > Several are intentional and demonstrate an understanding of two's > > complement arithmetic. > > Yes, but let me try to rephrase my last question to Dan: > > If an Impala user adds 1 to the maximum signed integer value, > should they expect that it will overflow to the minimum signed > integer value? If so, presumably we should have a test for that. Most of these overflows are not related to user values though, are they? I would expect most of these overflows to be on internal state. I think we should just look at them (or a representative sample) and see what makes sense. -- To view, visit http://gerrit.cloudera.org:8080/5082 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I88c7234bd7c5eb7404490a0913d90470c10835e7 Gerrit-PatchSet: 2 Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-Owner: Jim Apple <[email protected]> Gerrit-Reviewer: Dan Hecht <[email protected]> Gerrit-Reviewer: Jim Apple <[email protected]> Gerrit-Reviewer: Tim Armstrong <[email protected]> Gerrit-HasComments: No
