On 6 July 2016 at 09:30, David Woodhouse <[email protected]> wrote: > On Tue, 2016-07-05 at 17:04 +0000, Long, Qin wrote: >> Yes, this unset issue was already fixed in OpenSSL HEAD. >> The patch is OK for me to ignore the warning for current 1.0.2 >> version. Or we can backport some cleanups into our 1.0.2xx patch. > > My main concern is that we don't accumulate hacks and workarounds. So > add it to OpensslLib.inf by all means, but add a comment indicating the > specific issue (OpenSSL RT# under which it was fixed, and/or the commit > in OpenSSL HEAD Which fixed it), and that it's fixed in 1.1 so can be > dropped then. > > Otherwise this patch is just building up technical debt, because > *someone* later needs to clean these things up and make sure they're > not stale. Excessive warning-suppression in this code has *already* led > to subtle bugs which the warnings *would* have told us about. >
In general, I strongly agree with David's position here. However, I have encountered cases where RVCT's assign-before-reference diagnostics really get it wrong, and so I would not be opposed to disabling this particular warning globally (for RVCT). _______________________________________________ edk2-devel mailing list [email protected] https://lists.01.org/mailman/listinfo/edk2-devel
