Exhibit A: From #build on irc: <ted> well that's exciting: https://tbpl.mozilla.org/php/getParsedLog.php?id=13682580&tree=Try <ted> configure: error: Breakpad tools do not support compiling on WINNT while targeting WINNT. Use --disable-crashreporter. <ted> that's in a <ted> if (test "$OS_ARCH" != "$HOST_OS_ARCH"); then Opt build, slave w64-ix-slave51, try rev e8c115e0bda9. Rebuilds on w64-ix-slave54 & w64-ix-slave26 were able to configure. Exhibit B: https://tbpl.mozilla.org/php/getParsedLog.php?id=13758098&tree=Try for mounir. Debug build, slave w64-ix-slave31, try rev e1413f1c6566. A rebuild got past the error on w64-ix-slave34. These are the only two instances I can find in grepping over 14 days of try logs (opt & debug). Any ideas on what might be happening Ted ?
No. That IRC snippet was from my *own* try build, I was flummoxed. A rebuild succeeded, so it wasn't an error in the patches I pushed.
Not just try either: https://tbpl.mozilla.org/php/getParsedLog.php?id=13764979&tree=Mozilla-Inbound WINNT 5.2 mozilla-inbound build on 2012-07-23 00:21:12 PDT for push fda9279d7031 slave: w64-ix-slave80
Not actually just win32-on-win64, we've had them intermittently on win64 all along, we just haven't cared. https://tbpl.mozilla.org/php/getParsedLog.php?id=14176217&tree=Mozilla-Inbound
I have no idea what's going on here. Given the error message I don't see how the test can fail. The only thing that's remotely suspicious to me is that the test is wrapped in unnecessary parentheses, which means that it will be executed in a subshell. Perhaps there's some weirdness there? http://mxr.mozilla.org/mozilla-central/source/configure.in#5807
It would certainly not hurt to remove those parentheses.
Created attachment 652120 [details] [diff] [review] Remove parentheses Removes a couple of superfluous parentheses.