Bug 518506 comment 43 pointed out that an initialized declaration in the conditional of a for statement, added in that bug, confused VC7.1. http://hg.mozilla.org/mozilla-central/annotate/5a63989cada8/widget/src/xpwidgets/nsBaseWidget.cpp#l1005 AFAIK this is valid ISO C++, since at least 1998, but https://developer.mozilla.org/En/Developer_Guide/Build_Instructions/Windows_Prerequisites says that mozilla-central and mozilla-1.9.2 (should) build with VC7.1. (We don't support jemalloc with VC7.1, but that and VC8 may be the reason why jemalloc is off by default with msvc. http://hg.mozilla.org/mozilla-central/annotate/5a63989cada8//configure.in#l6713 ) Neil, do you know whether this is the only problem when compiling with VC7.1? I'm assuming this caused a compiler error. Do you know whether there is a good way to un-confuse VC7.1? I guess the only option is to move the declaration outside the loop? (Out of interest, I guess the same problem would happen with if, while, and switch statements?)
(In reply to comment #0) > Neil, do you know whether this is the only problem when compiling with VC7.1? Actually some of tracemonkey doesn't compile with VC7.1, although there is a pending review (for some time now, so probably bitrotted...) I can't remember any other problems offhand. > Do you know whether there is a good way to un-confuse VC7.1? > I guess the only option is to move the declaration outside the loop? That's what I did to get it to compile locally. > (Out of interest, I guess the same problem would happen with > if, while, and switch statements?) Probably.
Apparently VC7.1 understands initialized declarations in while-loop conditionals http://mxr.mozilla.org/mozilla-central/search?string=while+%28&find=&findi=&filter=while+\%28[*\w]%2B+\w%2B+%3D+&hitlimit=&tree=mozilla-central and maybe in if-statements http://mxr.mozilla.org/mozilla-central/search?string=if+%28&find=&findi=&filter=if+\%28[*\w]%2B+\w%2B+%3D+&hitlimit=&tree=mozilla-central e.g. http://hg.mozilla.org/mozilla-central/annotate/106fa9dc5896/testing/mochitest/ssltunnel/ssltunnel.cpp#l730 (I don't see any in switch-statements.) BTW, there is a similar initialized declaration in a for-loop conditional introduced after Neil's patches in bug 513909 http://hg.mozilla.org/mozilla-central/rev/ca95f2397bc5#l1.15
Created attachment 420442 [details] [diff] [review] xpwidget patch This is easy to fix, so let's fix. Other options are declaring otherRect (uninitialized) outside the for-loop, or declaring nextLink (initialized) outside a while-loop. But I like narrow variable scopes, so this dereferences twice (when not optimized) to declare otherRect inside the for-loop.
Blanket approval for Lorentz merge to mozilla-1.9.2 a=beltzner for 18.104.22.168 - please make sure to mark status1.9.2:.4-fixed
Merged into 1.9.2 at http://hg.mozilla.org/releases/mozilla-1.9.2/rev/84ba4d805430