> Date: Tue, 3 Jan 2006 03:11:46 -0500 > Cc: Alessandro Vesely <[EMAIL PROTECTED]>, [email protected] > From: "Paul D. Smith" <[EMAIL PROTECTED]> > > ez> Yes, but what feature of cmd.exe do we need for -j, and why? I > ez> need to know this in order to understand the problem and the > ez> solution, and in order to test the patched Make. > > ez> KJKHyperion says we need the caret-escaping, but it's not clear to me > ez> at all why this would allow -j. > > The critical thing that is needed from the shell to allow -jN to work is > that it can be invoked with open file descriptors (a pipe in this case) > and that it can further pass those open file descriptors to programs it > invokes, unmolested.
That is the default behavior on Windows (and, as a matter of fact, even on DOS). It was never any other way, because this part of file handling was copied from Unix back when MS-DOS v2 first introduced decent file I/O. > As far as I know that's all you need from the shell to get -jN working. Why ``from the shell''? is it because the shell might sometimes be used for command invocation (the ``slow'' method)? or for some other reasons, which are valid in the ``fast'' method as well? Anyway, if file handle inheritance is the only requirement on Posix systems that allows -jN, then the proposed patch does nothing in this area on Windows. I guess it is dealing with some Windows-specific aspect of Make, perhaps the invocation via batch files. I will have to dig into this when I have another slot of free time. Thanks for the explanations. _______________________________________________ Make-w32 mailing list [email protected] http://lists.gnu.org/mailman/listinfo/make-w32
