hans added a comment.
We tried it, and the warning is still firing in a similar (but not exactly the
same) way:
In file included from
../../content/browser/accessibility/browser_accessibility_manager_win.cc:19:
In file included from
../..\content/browser/renderer_host/legacy_render_widget_host_win.h:11:
../../third_party/wtl/include\atlapp.h(366,12): error: performing pointer
subtraction with a null pointer may have undefined behavior
[-Werror,-Wnull-pointer-subtraction]
uSize = NONCLIENTMETRICS_V1_SIZE;
^~~~~~~~~~~~~~~~~~~~~~~~
../../third_party/wtl/include\atlapp.h(248,38): note: expanded from macro
'NONCLIENTMETRICS_V1_SIZE'
#define NONCLIENTMETRICS_V1_SIZE _SIZEOF_STRUCT(NONCLIENTMETRICS,
lfMessageFont)
^~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
../../third_party/wtl/include\atlapp.h(228,91): note: expanded from macro
'_SIZEOF_STRUCT'
#define _SIZEOF_STRUCT(structname, member)
(((int)((LPBYTE)(&((structname*)0)->member) - ((LPBYTE)((structname*)0)))) +
sizeof(((structname*)0)->member))
^ ~~~~~~~~~~~~~~~~~~~~~~~~~~
Note that this time the warning is coming from a macro defined in a wtl/ header
itself , which is not a system header. So I think the "don't warn in system
headers" part is working correctly.
And it's good that it's got a separate flag, because for Chromium we would need
to turn it off.
But as a developer I'm confused by the warning saying this _may_ have undefined
behavior. Does it or doesn't it? If it does, I suppose the warning makes sense
and keeping it behind a separate flag that's part of -Wextra seems reasonable.
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D98798/new/
https://reviews.llvm.org/D98798
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
