The sole purpose of the TestBlockingProcess program is to live long enough to be able to kill it from the nsIProcess unit test. It currently relies on stdin to block, which is both a wrong assumption and an accurate way to have the nsIProcess unit test fail some particular environments (probably with the help of another bug in either nsProcess or NSPR, depending how one regards the issue, that I'll file right after this one). Such particular environments include those where stdin is not a tty (and it so happens in some Debian build environments, such as pbuilder). The result is that depending on the environment, TestBlockingProcess may or may not be "blocking". As its sole purpose is to be alive long enough to be killed by the nsIProcess unit test, it'd be just fine to loop on a sleep() call.

Backed out. e:/builds/moz2_slave/mozilla-central-win32/build/xpcom/tests/TestBlockingProcess.cpp(1) : fatal error C1083: Cannot open include file: 'unistd.h': No such file or directory http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1267790762.1267791982.4339.gz

(In reply to comment #3) > Backed out. > > e:/builds/moz2_slave/mozilla-central-win32/build/xpcom/tests/TestBlockingProcess.cpp(1) > : fatal error C1083: Cannot open include file: 'unistd.h': No such file or > directory > > http://tinderbox.mozilla.org/showlog.cgi?log=Firefox/1267790762.1267791982.4339.gz Really, I'm puzzled. unistd.h is a standard header, and it's used all over the place in the source... O_o

So, apparently, this should be protected in a #if !defined(XP_WIN) && !defined(XP_OS2) I don't know where sleep() is supposed to be defined on these.

It'd be better to use PR_Sleep, I guess. Let me try that.

This one should work everywhere