[PATCH] D98798: Produce warning for performing pointer arithmetic on a null pointer.

Wed May 12 10:40:56 PDT 2021

thakis added a comment.

In D98798#2754449 <https://reviews.llvm.org/D98798#2754449>, @jamieschmeiser wrote:

> @thakis, I have some questions about your comments.
> Are you sure this is coming from a system header?  The path that you gave has third_party as a directory in the path.

Yes. `third_party\depot_tools\win_toolchain\vs_files\20d5f2553f\Windows Kits\10\Include\10.0.19041.0\um\commctrl.h` is a hermetic system directory used via `/winsysrootthird_party\depot_tools\win_toolchain\vs_files\20d5f2553f`.

> If the warning were being triggered by code in a system header, I would have expected it to fire on something in the windows build bots but they appear to be clean.

I don't think LLVM uses a lot of code from the windows system headers. The compiler doesn't contain any UI :) (commctrl.h is a UI header).

> I don't know if it is possible to suppress a warning based on whether the source is in a system header, or from a macro expansion that is defined in a system header; are you aware about whether or not this is possible?

> If it is, I suspect it would already be in force as a general setting, again making me question whether this is a system header...and if it isn't a system header, that wouldn't help you in any case.
> If I understand your second point correctly, you have a system that previously compiled cleanly with warnings being treated as errors and you are concerned that if you use the existing options to suppress this particular warning, you are concerned that something could creep in.  Therefore you are proposing a different warning group that would be a subgroup of the existing one so that if you set it, you would set both but you would still be able to set the specific one.  I don't know if this is possible or not.  Either way, I suspect that it would require using a different warning for the subtraction case, which would also require significantly changing these changes. Am I understanding correctly?  If so, this implies that you have access to a workaround for your problem, although it may not be the best solution.
> I do not have access to a windows setup to test any of these proposed changes; in particular, given that I suspect that the affected files are specific to some third party vendor from which you have purchased code, I do not have means of investigating the actual problems/solutions.  If it would be helpful, I would be happy to review any changes that you might like to make to remedy the situation.
> I will be on vacation for the next few days so please excuse my delayed responses.

Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D98798/new/

https://reviews.llvm.org/D98798