This warning showed up recently on my own builds. Doesn't seem to be showing up on Tinderbox. I do PGO builds with MSVC 2010 with -O2 -GL. c:\mozbuild\mozilla-central\layout\style\nscssscanner.cpp(736) : warning C4789: destination of memory copy is too small

http://mxr.mozilla.org/mozilla-central/source/layout/style/nsCSSScanner.cpp?mark=736#721 721 PRBool 722 nsCSSScanner::Next(nsCSSToken& aToken) ... 734 // IDENT 735 if (StartsIdent(ch, Peek())) 736 return ParseIdent(ch, aToken);

This seems to have been tied to the optimization settings I was using. Whatever the reason, it's gone now.

I still see this on MSVC 2010 PGO builds with no optimizations enabled. This looks potentially bad. Maybe a bug with MSVC, but is there a way to mitigate it?

What line is the trunk warning on?

726 now. http://mxr.mozilla.org/mozilla-central/source/layout/style/nsCSSScanner.cpp?mark=726#711

Erm. Line 726 shouldn't have any memory copies....

Dunno, but I just checked again. c:\mozbuild\mozilla-central\layout\style\nscssscanner.cpp(726) : warning C4789: destination of memory copy is too small

If you inline ParseIdent at that point manually, which line does it complain on?

Here's what I'm trying now. Will post back in a couple hours with results.

Now it's giving the warning on this line: if (!GatherIdent(ch, ident)) {

Alright. Can you try inlining that too?

This is what I'm testing now. I'm pretty sure I did it right, but I would appreciate if you could give it a quick look to make sure I didn't screw something up.

Comment on attachment 583987 [details] [diff] [review] Inline ParseIdent and GatherIdent Yeah, looks reasonable.

else if (0 < ch) { >>> ident.Append(ch); }

Hrm. And this is a _compile_ time warning???

No, while linking xul.lib on the second pass.

Is there anything more I can do to help move this along?

I have no idea what's going on here. :( In general, our string append code should.. kinda work. We use it all over....

This also happens with tinderbox PGO builds now that they've been switched over to using VC10. https://tbpl.mozilla.org/php/getParsedLog.php?id=9016860&tree=Firefox&full=1

At the risk of shouting more into the Grand Canyon, can someone at least say if this bug is just a nuisance or something worse? This is shipping in nightlies now. From MSDN: Warns about buffer overrun when specific C run-time (CRT) functions are used, parameters are passed, and assignments are performed, such that the data sizes are known at compile time. This warning is for situations that might elude typical data-size mismatch detection.

That's the problem. I wish I could tell what MSVC _thinks_ is going on here.... :(