Re: OpenCOLLADA need a little c++ help
This is case in point for why you should not use -Werror when building
packages. Instead of adding pragmas in various places, I would track
down where the -Werror gets added and patch that out instead.
(Exception: it does make sense to use -Werror=format-security,
-Werror=implicit-function-declaration, etc. for specific warnings that
are unlikely to result in false positives. But using plain -Werror to
turn all warnings into errors is a really bad idea. Save that for
developer builds only; it's just not appropriate for release builds.)
On Mon, May 29 2023 at 09:26:20 AM -0400, Sam Varshavchik
<mrsam(a)courier-mta.com> wrote:
This all presumes that this is a false positive, which I'll
estimate
will be the case more often than not.
This particular warning has a very high rate of false positives, so
much so that you should strongly consider using
-Wno-dangling-reference. In my personal experience, the false positive
rate is 100%. It's too bad, because if there really *is* a dangling
reference, that would be a serious memory safety bug, but honestly I'm
starting to think it's not worth spending the time to investigate the
warning anymore. As for this particular case, it's returning a cached
XmlNode (cached in XmlDoc.mXPathCache), not a local variable, so it's
not a bug. Or at least sure looks like it's not a bug. :)
Michael